Artificial intelligence is rapidly changing many aspects of how we work and live. (How many stories did you read last week about self-driving cars and job-stealing robots? Perhaps your holiday shopping involved some AI algorithms, as well.) But despite the constant flow of news, many misconceptions about AI remain.
AI doesn’t think in our sense of the word at all, Scriffignano explains. “In many ways, it’s not really intelligence. It’s regressive.”
IT leaders should make deliberate choices about what AI can and can’t do on its own. “You have to pay attention to giving AI autonomy intentionally and not by accident,”
6,957 Comments
Tomi Engdahl says:
Can ChatGPT Create Web Design?
https://www.elegantthemes.com/blog/design/can-chatgpt-create-web-design
We know ChatGPT does a great job at writing text and giving you answers, but can it create web design? And if yes, to what extent? What are its limitations, and what does it particularly excel at? In this post, we’ll give you an overview of ChatGPTs capabilities and provide you with other AI options to build your next website!
So ChatGPT Can Design Websites… Right? Let’s give credit where it’s due — ChatGPT can do some pretty impressive things with web design. But before we hand over our design credentials to AI, let’s see what it can handle. Planning Your Site Structure Drop ChatGPT a line about your business, and it’ll sketch out a basic page list. Home page? Check. About page? Obviously. But it goes deeper,
Generating Code For Your Website Need a navigation menu in HTML? Or maybe a simple contact form? ChatGPT spits out working code snippets that do the job. We’ve tested dozens of its code blocks — most work straight out of the box, though you might need to tweak a class name here or there. The funny thing is, it’s better at fixing code than writing it from scratch. Paste in your broken CSS, and it’ll spot that missing semicolon faster than your tired eyes ever would. However, watch out when it gets too confident.
Quick Fixes And Simple Scripts Sometimes, you just need a quick JavaScript function to validate a phone number or a snippet to make that sticky header stick. ChatGPT handles these small coding tasks surprisingly well, especially when you need straightforward solutions to common problems. It’s good at generating form validation scripts, basic animations, and those little interactive elements that make websites feel more polished. The code it provides usually comes with explanations, making it easier to understand what each part does before implementing it.
Handy Design Suggestions Regarding the visual side of things, ChatGPT can help brainstorm design elements that make sense. Ask it about hero section layouts, and it’ll suggest ways to arrange your content that follow solid design principles. Want ideas for your service cards? It’ll outline different approaches to showcase your offerings.
Coming Up With Great Copy Need words that sell? ChatGPT can help spark ideas for headlines, button text, and tricky sections where you’re drawing a blank. It’s particularly useful when you need multiple versions of the same message — like turning “We build websites” into ten different ways to say it without sounding repetitive.
But, Here’s Where ChatGPT Falls Short Reality check time. Those viral success stories and perfectly edited YouTube tutorials? They’re not telling you everything. The internet is busy celebrating AI’s web design revolution, but some crucial details are getting swept under the rug. Let’s pull back the curtain. Perfecting Prompts — Your New Full-Time Job Can you just ask ChatGPT to design your website and call it a day? Not quite. Getting usable results means learning a new skill — crafting the perfect prompt.
No Preview Means Endless Tweaking Here’s the thing about working with ChatGPT — it’s all text. You ask for a design, and it responds like a friend describing your dream house through a game of 20 questions. “Make the header blue” sounds simple enough, but which blue? Navy? Sky? Is that weird blue-green trending right now?
You’ll find yourself in this loop: get code, paste it in, check the result, return to ChatGPT, tweak the prompt, get new code, paste it again… rinse and repeat. That button you wanted to be centered? It’s now floating somewhere near Saturn. The “subtle shadow” it suggested? Looks more like your element is having an identity crisis.
You’re playing design ping-pong without a visual preview between your code editor and ChatGPT. What could be a simple drag-and-drop in a visual editor becomes a game of trial and error that would test the patience of a saint.
Good Luck With Mobile Responsiveness Getting your site to look good on mobile is already tricky — now try doing it through text prompts alone. ChatGPT might give you perfectly functional desktop code, but mobile? That’s where things get messy. Your beautifully arranged desktop layout suddenly looks like a game of Tetris gone wrong on phones.
Updates? Hope You Saved Those Prompts Remember that perfect hero section ChatGPT helped you create last month? Now, you want to tweak it. Small problem: you didn’t save the original prompt. You’re back to square one, trying to recreate the same code while maintaining consistency with the rest of your site’s style.
This is where ChatGPT’s stateless nature becomes a real pain. Each conversation starts fresh, with zero context from your previous design decisions. That means no version control, no easy updates, and definitely no “just change the color” quick fixes. Good luck getting the same output twice, even if you saved the prompt.
What should be a five-minute update turns into an hour of prompt archaeology, trying to remember exactly how you worded things to get that perfect result. And with each new attempt, you risk ending up with slightly different code that breaks your site’s consistency. When Code Breaks (And It Will) So your AI-generated code has been running smoothly for weeks, and then suddenly, boom. Something breaks. Maybe it’s a browser update, or that fancy animation conflicts with another script.
Debugging becomes a game of asking ChatGPT to fix issues in code it generated but doesn’t remember generating. Each fix potentially introduces new problems, and you’re caught in an endless troubleshooting cycle.
Beyond ChatGPT: AI Web Design Capabilities In Divi WordPress powers over 43% of all websites, making it the world’s most trusted platform for web design. Its open-source nature and flexible architecture let you build anything from simple blogs to complex online stores with complete design freedom. WordPress becomes a powerhouse when paired with the right tools. Divi amplifies WordPress’s capabilities by combining intelligent design features, extensive templates, and AI-driven solutions in one seamless package.
Tomi Engdahl says:
Divi is the Theme and Page Builder combo that brings simple no-code editing, advanced design systems, and powerful AI tools to WordPress.
https://www.elegantthemes.com/gallery/divi/
Tomi Engdahl says:
Noteworthy AI Research Papers of 2024 (Part One)
Six influential AI papers from January to June
https://magazine.sebastianraschka.com/p/ai-research-papers-2024-part-1
Tomi Engdahl says:
3 Ideas To Make $1,000+ This Month With ChatGPT
https://www.forbes.com/sites/rachelwells/2024/12/29/3-ideas-to-make-1000-this-month-with-chatgpt/
Can you imagine making an extra $1,000 or more a month on top of your regular salary? A recent Bankrate study revealed that the average American needs to earn about $186,000 a year to live reasonably comfortably; yet the average U.S. salary is approximately 42% of this figure—$79,000, a far cry from what most Americans need to live.
This is one major factor driving the surge of side hustles and the gig economy. And thanks to technology and recent developments in AI, starting a highly profitable side hustle has never been easier. When you have tools like ChatGPT at your fingertips, you have one of your best partners to hit and exceed your financial goals. And what’s more, ChatGPT not only unlocks your earning potential, but more than this, it helps you leverage the skills and knowledge you already have and expands on them.
3 ChatGPT Business Ideas To Make $1,000+ A Month In 2025
Are you ready to start making more money every month so you can afford necessities, shrink your debt, invest, or have extra discretionary income just because? Check out the below ChatGPT side hustle ideas:
1. Be A Freelance Ad Creative
A freelance ad creative assists businesses by creating the elements of their marketing campaign including ad copy; and if your skills are versatile enough, you can even assist with creating the brand’s campaign strategy for their product, brand, or service.
Ad creatives on Upwork typically tend to charge between $60 to $115 per hour for their services, which include commercial video design, graphic design, social media content creation, development of brand identity and guidelines, and even keyword research.
2. Offer Freelance Event Planning Services
Another fantastic ChatGPT-powered side hustle is event planning. You can even make it more niched and focus on virtual or online event planning for businesses. To speed up the process, you can give ChatGPT prompts such as:
Suggest a detailed event itinerary for an online conference for [type of business]. The event’s theme is [name of theme] and there will be XXX attendees.
Freelance event planners typically charge based on a percentage of a total event budget, which means you can easily see yourself making a little under $10,000 and as much as even up to $50,000 for a large-scale event with your wealth of experience.
3. ChatGPT-Powered App Development
As a freelance developer you can create chatbots and customized applications for small businesses who are seeking to leverage the power of artificial intelligence to streamline their operations. Some developers charge $250 per hour for an app development project, making it quite easy to hit the $1,000 a month target with just one client or project.
For instance, using ChatGPT Canvas you can test the code you’re writing, and you can use o1 in ChatGPT Pro (the advanced $200-a-month subscription) for complex analysis, debugging, and troubleshooting of your code to help you refine and edit your application.
Tomi Engdahl says:
How Generative AI And LLMs Are Reinventing Our Vocabulary Such That We Might Lose Our Grasp On Human Languages
https://www.forbes.com/sites/lanceeliot/2024/12/29/how-generative-ai-and-llms-are-reinventing-our-vocabulary-such-that-we-might-lose-our-grasp-on-human-languages/
Tomi Engdahl says:
What is an AI agent? A computer scientist explains the next wave of artificial intelligence tools
https://www.rappler.com/technology/features/what-is-ai-agent-explainer-artificial-intelligence-tools/
Tomi Engdahl says:
I put ChatGPT vs Claude to the test with 7 prompts — here’s the winner
https://www.tomsguide.com/ai/chatgpt/i-put-chatgpt-vs-claude-to-the-test-with-7-prompts-heres-the-winner
Tomi Engdahl says:
https://www.rappler.com/technology/features/what-is-ai-agent-explainer-artificial-intelligence-tools/
Tomi Engdahl says:
OpenAI CEO Sam Altman on Sunday said that the company is currently losing money on its $200-per-month ChatGPT Pro plan because people are using it more than the company expected.
ChatGPT Pro, launched late last year, grants access to an upgraded version of OpenAI’s o1 “reasoning” AI model, o1 pro mode, and lifts rate limits on several of the company’s other tools, including its Sora video generator.
Read more from Kyle Wiggers on OpenAI losing money on ChatGPT Pro here: https://tcrn.ch/4gJrXHy
#TechCrunch #technews #artificialintelligence #OpenAI #ChatGPT #SamAltman
Tomi Engdahl says:
Nvidia is getting into world models — AI models that take inspiration from the mental models of the world that humans develop naturally.
At #CES2025 in Las Vegas, the company announced that it is making openly available a family of world models that can predict and generate “physics-aware” videos. Nvidia is calling this family Cosmos World Foundation Models, or Cosmos WFMs for short.
Read more from Kyle Wiggers here: https://tcrn.ch/3DMtAWn
#TechCrunch #technews #artificialintelligence #Nvidia #JensenHuang
Tomi Engdahl says:
https://www.youtube.com/@Green-Code
I Built a Neural Network from Scratch
https://www.youtube.com/watch?v=cAkMcPfY_Ns
AI Learns What Pizza Is
https://www.youtube.com/watch?v=2xMLlm_VDJE
AI Learns the Numbers
https://www.youtube.com/watch?v=2utAfvGAbgg
Tomi Engdahl says:
Sean Hollister / The Verge:
Razer unveils Project Ava, an “AI gaming copilot” that helps gamers by taking thousands of pictures of their screen and telling them how to play in real time
Razer Project Ava: would you pay an AI to help you get good at games?
/ Bit of a strange fit for the Razer brand — particularly given how intrusive the prototype can be.
https://www.theverge.com/24338016/razer-project-ava-ai-gaming-copilot-coach
Gaming hardware manufacturer Razer is introducing what might be its most controversial “project” yet — an “AI gaming copilot” designed to help you get better at video games. With your permission, it takes thousands of pictures of your screen, then tells you how to play a game, optimally in real time, as you continue to play.
“By analyzing millions of simulations within seconds, I’ll always have an answer to your toughest gaming challenges,” a Razer marketing video claims.
Tomi Engdahl says:
6 AI trends you’ll see more of in 2025
https://news.microsoft.com/source/features/ai/6-ai-trends-youll-see-more-of-in-2025/?ocid=pd_spnsrd-post_techmeme_jan-25_6-ai-trends
In 2025, AI will evolve from a tool for work and home to an integral part of both.
AI-powered agents will do more with greater autonomy and help simplify your life at home and on the job. On the global stage, AI will help us find new ways to address some of the biggest challenges we face, from the climate crisis to healthcare access.
This progress will be driven by advancements in AI’s ability to remember more and reason better, among other innovations. And Microsoft will remain grounded in its commitment to help people use and build AI that is safe and secure.
“AI is already making the impossible feel possible, and over the past year we’ve seen significant numbers of people and organizations moving from AI experimentation to more meaningful adoption,” says Chris Young, executive vice president of business development, strategy and ventures at Microsoft. “This is the start of a full-scale transformation of how this technology will change every part of our lives.”
In the last year alone, generative AI usage jumped from 55% to 75% among business leaders and AI decisionmakers. New AI tools will bring even more potential.
Want to know what’s ahead? Here are six AI trends to watch — and how Microsoft will innovate on each — in 2025.
Over the past year, AI models became faster and more efficient. Today, large-scale “frontier models” can complete a broad range of tasks from writing to coding, and highly specialized models can be tailored for specific tasks or industries.
In 2025, models will do more — and do it even better.
Models with advanced reasoning capabilities, like OpenAI o1, can already solve complex problems with logical steps that are similar to how humans think before responding to difficult questions. These capabilities will continue to be useful in fields like science, coding, math, law and medicine, allowing models to compare contracts, generate code and execute multistep workflows.
Making models faster, better and more specialized will create new and more useful AI experiences, including with agents, in 2025.
Workers at nearly 70% of Fortune 500 companies already use Microsoft 365 Copilot to tackle plenty of repetitive and mundane tasks, such as sifting through email and taking notes during Teams meetings. In 2025, a new generation of AI-powered agents will do more — even handling certain tasks on your behalf.
“Think of agents as the apps of the AI era,” says Charles Lamanna, corporate vice president of business and industry Copilot. “Just as we use different apps for various tasks, agents will begin to transform every business process, revolutionizing the way we work and manage our organizations.”
With advancements in memory, reasoning and multimodal capabilities, agents will handle more complex assignments with new skills and ways to interact.
Organizations can reimagine processes like creating reports and human resources tasks such as helping resolve a new laptop issue or answering benefits questions, freeing employees to focus on higher value work.
And you can build and use agents no matter your technical skill. Anyone can build an agent in Copilot Studio — no coding required — while developers can create more sophisticated agents to orchestrate more complex tasks in Azure AI Foundry.
All of this will lay the groundwork for a future when organizations have a constellation of agents — from simple prompt-and-response to fully autonomous — that will work independently or together on behalf of individuals, groups or functions to execute and orchestrate processes.
Amid all this AI development, human oversight will remain a central cog in the evolving AI-powered agent wheel, says Kamar.
“In 2025, a lot of conversation will be about drawing the boundaries around what agents are allowed and not allowed to do, and always having human oversight,” Kamar says.
Tomi Engdahl says:
What are AI agents?
https://www.ibm.com/think/topics/ai-agents
An artificial intelligence (AI) agent refers to a system or program that is capable of autonomously performing tasks on behalf of a user or another system by designing its workflow and utilizing available tools.
AI agents can encompass a wide range of functionalities beyond natural language processing including decision-making, problem-solving, interacting with external environments and executing actions.
These agents can be deployed in various applications to solve complex tasks in various enterprise contexts from software design and IT automation to code-generation tools and conversational assistants. They use the advanced natural language processing techniques of large language models (LLMs) to comprehend and respond to user inputs step-by-step and determine when to call on external tools.
How AI agents work
At the core of AI agents are large language models (LLMs). For this reason, AI agents are often referred to as LLM agents. Traditional LLMs, such as IBM® Granite™ models, produce their responses based on the data used to train them and are bounded by knowledge and reasoning limitations. In contrast, agentic technology uses tool calling on the backend to obtain up-to-date information, optimize workflow and create subtasks autonomously to achieve complex goals.
In this process, the autonomous agent learns to adapt to user expectations over time. The agent’s ability to store past interactions in memory and plan future actions encourages a personalized experience and comprehensive responses.
1 This tool calling can be achieved without human intervention and broadens the possibilities for real-world applications of these AI systems. The approach that AI agents take in achieving goals set by users is comprised of these three stages:
Goal initialization and planning
Although AI agents are autonomous in their decision-making processes, they require goals and environments defined by humans.2 There are three main influences on autonomous agent behavior:
The team of developers that design and train the agentic AI system.
The team that deploys the agent and provides the user with access to it.
The user that provides the AI agent with specific goals to accomplish and establishes available tools to use.
Given the user’s goals and the agent’s available tools, the AI agent then performs task decomposition to improve performance.3 Essentially, the agent creates a plan of specific tasks and subtasks to accomplish the complex goal.
For simple tasks, planning is not a necessary step. Instead, an agent can iteratively reflect on its responses and improve them without planning its next steps.
Reasoning using available tools
AI agents base their actions on the information they perceive. Often, AI agents do not have the full knowledge base needed for tackling all subtasks within a complex goal. To remedy this, AI agents use their available tools. These tools can include external data sets, web searches, APIs and even other agents. After the missing information is retrieved from these tools, the agent can update its knowledge base. This means that each step of the way, the agent reassesses its plan of action and self-corrects.
The agent can now combine the information it has learned from its tools to identify patterns.
Learning and reflection
AI agents use feedback mechanisms, such as other AI agents and human-in-the-loop (HITL), to improve the accuracy of their responses. Let’s return to our previous surfing example to highlight this. After the agent forms its response to the user, the agent stores the learned information along with the user’s feedback to improve performance and adjust to user preferences for future goals.
If other agents were used to reach the goal, their feedback may also be used. Multi-agent feedback can be especially useful in minimizing the time that human users spend providing direction. However, users can also provide feedback throughout the agent’s actions and internal reasoning to better align the results with the intended goal.2
Feedback mechanisms improve the AI agent’s reasoning and accuracy, which is commonly referred to as iterative refinement.3 To avoid repeating the same mistakes, AI agents can also store data about solutions to previous obstacles in a knowledge base.
Agentic versus non-agentic AI chatbots
AI chatbots use conversational AI techniques such as natural language processing (NLP) to understand user questions and automate responses to them. These chatbots are a modality whereas agency is a technological framework.
Non-agentic AI chatbots are ones without available tools, memory and reasoning. They can only reach short-term goals and cannot plan ahead. As we know them, non-agentic chatbots require continuous user input to respond. They can produce responses to common prompts that most likely align with user expectations but perform poorly on questions unique to the user and their data. Since these chatbots do not hold memory, they cannot learn from their mistakes if their responses are unsatisfactory.
In contrast, agentic AI chatbots learn to adapt to user expectations over time, providing a more personalized experience and comprehensive responses. They can complete complex tasks by creating subtasks without human intervention and considering different plans. These plans can also be self-corrected and updated as needed. Agentic AI chatbots, unlike non-agentic ones, assess their tools and use their available resources to fill in information gaps.
Reasoning paradigms
There is not one standard architecture for building AI agents. Several paradigms exist for solving multi-step problems.
ReAct (Reasoning and Action)
With this paradigm, we can instruct agents to “think” and plan after each action taken and with each tool response to decide which tool to use next. These Think-Act-Observe loops are used to solve problems step by step and iteratively improve upon responses.
Through the prompt structure, agents can be instructed to reason slowly and to display each “thought”.4 The agent’s verbal reasoning gives insight into how responses are formulated. In this framework, agents continuously update their context with new reasoning. This can be interpreted as a form of Chain-of-Thought prompting.
ReWOO (Reasoning WithOut Observation)
The ReWOO method, unlike ReAct, eliminates the dependence on tool outputs for action planning. Instead, agents plan upfront. Redundant tool usage is avoided by anticipating which tools to use upon receiving the initial prompt from the user. This is desirable from a human-centered perspective since the user can confirm the plan before it is executed.
The ReWOO workflow is made up of three modules. In the planning module, the agent anticipates its next steps given a user’s prompt. The next stage entails collecting the outputs produced by calling these tools. Lastly, the agent pairs the initial plan with the tool outputs to formulate a response. This planning ahead can greatly reduce token usage and computational complexity as well as the repercussions of intermediate tool failure.
Types of AI agents
AI agents can be developed to have varying levels of capabilities. A simple agent may be preferred for straightforward goals to limit unnecessary computational complexity. In order of simplest to most advanced, there are 5 main agent types:
1. Simple reflex agents
Simple reflex agents are the simplest agent form that grounds actions on current perception. This agent does not hold any memory, nor does it interact with other agents if it is missing information. These agents function on a set of so-called reflexes or rules. This means that the agent is preprogrammed to perform actions that correspond to certain conditions being met.
If the agent encounters a situation that it is not prepared for, it cannot respond appropriately.
2. Model-based reflex agents
Model-based reflex agents use both their current perception and memory to maintain an internal model of the world. As the agent continues to receive new information, the model is updated. The agent’s actions depend on its model, reflexes, previous precepts and current state.
These agents, unlike simple reflex agents, can store information in memory and can operate in environments that are partially observable and changing. However, they are still limited by their set of rules.6
Example: A robot vacuum cleaner. As it cleans a dirty room, it senses obstacles such as furniture and adjusts around them. The robot also stores a model of the areas it has already cleaned to not get stuck in a loop of repeated cleaning.
3. Goal-based agents
Goal-based agents have an internal model of the world and also a goal or set of goals. These agents search for action sequences that reach their goal and plan these actions before acting on them. This search and planning improve their effectiveness when compared to simple and model-based reflex agents.7
Example: A navigation system that recommends the fastest route to your destination.
4. Utility-based agents
Utility-based agents select the sequence of actions that reach the goal and also maximize utility or reward. Utility is calculated using a utility function. This function assigns a utility value, a metric measuring the usefulness of an action or how “happy” it will make the agent, to each scenario based on a set of fixed criteria.
The criteria can include factors such as progression toward the goal, time requirements, or computational complexity. The agent then selects the actions that maximize the expected utility. Hence, these agents are useful in cases where multiple scenarios achieve a desired goal and an optimal one must be selected.7
Example: A navigation system that recommends the route to your destination that optimizes fuel efficiency and minimizes the time spent in traffic and the cost of tolls. This agent measures utility through this set of criteria to select the most favorable route.
5. Learning agents
Learning agents hold the same capabilities as the other agent types but are unique in their ability to learn. New experiences are added to their initial knowledge base, which occurs autonomously. This learning enhances the agent’s ability to operate in unfamiliar environments. Learning agents may be utility or goal-based in their reasoning and are comprised of four main elements:7
Learning: This improves the agent’s knowledge by learning from the environment through its precepts and sensors.
Critic: This provides feedback to the agent on whether the quality of its responses meets the performance standard.
Performance: This element is responsible for selecting actions upon learning.
Problem generator: This creates various proposals for actions to be taken.
Example: Personalized recommendations on e-commerce sites. These agents track user activity and preferences in their memory. This information is used to recommend certain products and services to the user.
Benefits of AI agents
Task automation
With the ongoing advancements in generative AI, there is a growing interest in workflow optimization using AI, or intelligent automation. AI agents are AI tools that can automate complex tasks that would otherwise require human resources.
Greater performance
Multi-agent frameworks tend to outperform singular agents.11 This is because the more plans of action are available to an agent, the more learning and reflection occur. An AI agent incorporating knowledge and feedback from other AI agents specializing in related areas can be useful for information synthesis.
Quality of responses
AI agents provide responses that are more comprehensive, accurate and personalized to the user than traditional AI models. This is extremely important to us as users since higher-quality responses typically yield a better customer experience. As previously described, this is made possible through exchanging information with other agents, using external tools and updating their memory stream. These behaviors emerge on their own and are not preprogrammed.
Risks and limitations
Multi-agent dependencies
Certain complex tasks require the knowledge of multiple AI agents. When implementing these multi-agent frameworks, there is a risk of malfunction. Multi-agent systems built on the same foundation models may experience shared pitfalls. Such weaknesses could cause a system-wide failure of all involved agents or expose vulnerability to adverse attacks.
Infinite feedback loops
The convenience of the hands-off reasoning for human users using AI agents also comes with its risks. Agents that are unable to create a comprehensive plan or reflect on their findings, may find themselves repeatedly calling the same tools, invoking infinite feedback loops. To avoid these redundancies, some level of real-time human monitoring may be used.1
Computational complexity
Building AI agents from scratch is both time-consuming and can also be very computationally expensive. The resources required for training a high-performance agent can be extensive. Additionally, depending on the complexity of the task, agents can take several days to complete tasks.
Best practices
Activity logs
To address the concerns of multi-agent dependencies, developers can provide users with access to a log of agent actions.14 The actions can include the use of external tools and describe the external agents utilized to reach the goal. This transparency grants users insight into the iterative decision-making process, provides the opportunity to discover errors and builds trust.
Interruption
Preventing AI agents from running for overly long periods of time is recommended. Particularly, in cases of unintended infinite feedback loops, changes in access to certain tools, or malfunctioning due to design flaws. One way to accomplish this is by implementing interruptibility.
Maintaining control of this involves allowing human users the option to gracefully interrupt a sequence of actions or the entire operation. Choosing if and when to interrupt an AI agent requires some thoughtfulness as some terminations can cause more harm than good. For instance, it may be safer to allow a faulty agent to continue assisting in a life-threatening emergency than to completely shut it down.5
Unique agent identifiers
To mitigate the risk of agentic systems being used for malicious use, unique identifiers can be used.14 If these identifiers were to be required for agents to access external systems, there would be greater ease in tracing the origin of the agent’s developers, deployers and its user. This would be particularly helpful in case of any malicious use or unintended harm done by the agent. This level of accountability would provide a safer environment for these AI agents to operate.
Human supervision
To assist in the learning process for AI agents, especially in their early stages in a new environment, it can be helpful to provide occasional human feedback. This allows the AI agent to compare its performance to the expected standard and adjust accordingly. This form of feedback is helpful in improving the agent’s adaptability to user preferences.5
Apart from this, it is best practice to require human approval before an AI agent takes highly impactful actions. For instance, actions ranging from sending mass emails to financial trading should require human confirmation.7 Some level of human monitoring is recommended for such high-risk domains.
Tomi Engdahl says:
Michael Acton / Financial Times:
Jensen Huang predicts robotics could become “the largest technology industry the world has ever seen”; Nvidia: the humanoids market will reach $38B in 20 years — Jensen Huang unveils AI models for humanoids and self-driving car partnership with Toyota at CES keynote
Nvidia chief calls robots ‘multitrillion-dollar’ opportunity
Jensen Huang unveils AI models for humanoids and self-driving car partnership with Toyota at CES keynote
https://www.ft.com/content/a8ce27a1-bd4c-4bc5-bec6-8dfcdc7d4d31
Tomi Engdahl says:
Dean Takahashi / VentureBeat:
Nvidia debuts the Llama Nemotron and Cosmos Nemotron family of models to advance agentic AI, in Nano, Super, and Ultra sizes, on Nvidia’s site and Hugging Face — Nvidia announced Nemotron Model Families to advance agentic AI as part of its bevy of AI announcements at CES 2025 today.
Nvidia’s Nemotron Model Families will advance AI agents
https://venturebeat.com/ai/nvidias-nemotron-model-families-will-advance-ai-agents/
Tomi Engdahl says:
Dean Takahashi / VentureBeat:
Nvidia unveils Mega, an “Omniverse Blueprint” for developing, testing, and optimizing physical AI and robot fleets at scale in a digital twin before deployment
Nvidia unveils Mega Omniverse blueprint for building industrial robot fleet digital twins
https://venturebeat.com/ai/nvidia-unveils-mega-omniverse-blueprint-for-building-industrial-robot-fleet-digital-twins/
Tomi Engdahl says:
Will Knight / Wired:
Jensen Huang says Nvidia trained Cosmos models for humanoids, robots, and cars on 20M hours of footage of “humans walking, hands moving, manipulating things”
Nvidia’s ‘Cosmos’ AI Helps Humanoid Robots Navigate the World
Nvidia CEO Jensen Huang says the new family of foundational AI models was trained on 20 million hours of “humans walking, hands moving, manipulating things.”
https://www.wired.com/story/nvidia-cosmos-ai-helps-robots-self-driving-cars/
Tomi Engdahl says:
Dean Takahashi / VentureBeat:
Nvidia gives select developers early access to Omniverse Cloud Sensor RTX, enabling smarter autonomous machines with generative AI, via APIs supporting sensors
Nvidia announces early access for Omniverse Sensor RTX for smarter autonomous machines
https://venturebeat.com/ai/nvidia-announces-early-access-for-omniverse-sensor-rtx-for-smarter-autonomous-machines/
Nvidia announced early access for Omniverse Cloud Sensor RTX software to enable smarter autonomous machines with generative AI.
Generative AI and foundation models let autonomous machines generalize beyond the operational design domains on which they’ve been trained. Using new AI techniques such as tokenization and large language and diffusion models, developers and researchers can now address longstanding hurdles to autonomy. Nvidia made the announcement during Nvidia CEO Jensen Huang’s keynote at CES 2025.
These larger models require massive amounts of diverse data for training, fine-tuning and validation. But collecting such data — including from rare edge cases and potentially hazardous scenarios, like a pedestrian crossing in front of an autonomous vehicle (AV) at night or a human entering a welding robot work cell — can be incredibly difficult and resource-intensive.
To help developers fill this gap, NVIDIA Omniverse Cloud Sensor RTX APIs enable physically accurate sensor simulation for generating datasets at scale. The application programming interfaces (APIs) are designed to support sensors commonly used for autonomy — including cameras, radar and lidar — and can integrate seamlessly into existing workflows to accelerate the development of autonomous vehicles and robots of every kind.
Omniverse Sensor RTX APIs are now available to select developers in early access. Organizations such as Accenture, Foretellix, MITRE and Mcity are integrating these APIs via domain-specific blueprints to provide end customers with the tools they need to deploy the next generation of industrial manufacturing robots and self-driving cars.
Powering Industrial AI With Omniverse Blueprints
In complex environments like factories and warehouses, robots must be orchestrated to safely and efficiently work alongside machinery and human workers. All those moving parts present a massive challenge when designing, testing or validating operations while avoiding disruptions.
Mega is an Omniverse Blueprint that offers enterprises a reference architecture of Nvidia accelerated computing, AI, NVIDIA Isaac and NVIDIA Omniverse technologies.
Building Smarter Autonomous Machines: NVIDIA Announces Early Access for Omniverse Sensor RTX
Organizations including Accenture and Foretellix are accelerating the development of next-generation self-driving cars and robots with high-fidelity, scalable sensor simulation.
https://blogs.nvidia.com/blog/omniverse-sensor-rtx-autonomous-machines/
Share
Generative AI and foundation models let autonomous machines generalize beyond the operational design domains on which they’ve been trained. Using new AI techniques such as tokenization and large language and diffusion models, developers and researchers can now address longstanding hurdles to autonomy.
These larger models require massive amounts of diverse data for training, fine-tuning and validation. But collecting such data — including from rare edge cases and potentially hazardous scenarios, like a pedestrian crossing in front of an autonomous vehicle (AV) at night or a human entering a welding robot work cell — can be incredibly difficult and resource-intensive.
To help developers fill this gap, NVIDIA Omniverse Cloud Sensor RTX APIs enable physically accurate sensor simulation for generating datasets at scale. The application programming interfaces (APIs) are designed to support sensors commonly used for autonomy — including cameras, radar and lidar — and can integrate seamlessly into existing workflows to accelerate the development of autonomous vehicles and robots of every kind.
Omniverse Sensor RTX APIs are now available to select developers in early access. Organizations such as Accenture, Foretellix, MITRE and Mcity are integrating these APIs via domain-specific blueprints to provide end customers with the tools they need to deploy the next generation of industrial manufacturing robots and self-driving cars.
Tomi Engdahl says:
Jay Peters / The Verge:
Nvidia updates its Autonomous Game Characters with small language models, saying the NPCs can “perceive, plan, and act like human players” in games like PUBG
Nvidia’s AI NPCs are no longer chatbots — they’re your new PUBG teammate
/ PUBG is getting AI teammates powered by Nvidia technology.
https://www.theverge.com/2025/1/6/24337949/nvidia-ace-ai-npcs-pubg-ally-teammate
Tomi Engdahl says:
Reuters:
Satya Nadella says Microsoft will invest about $3B to expand its Azure cloud and AI capacity in India and train 10 million people in AI in the country by 2030 — Microsoft (MSFT.O) will invest about $3 billion to expand capacity for artificial intelligence and its Azure cloud-computing services in India …
Microsoft to invest $3 billion in India, to expand AI and cloud capacity
https://www.reuters.com/technology/microsoft-invest-3-bln-expand-azure-ai-capacity-india-2025-01-07/
Tomi Engdahl says:
OpenAI’s Latest AI Can Cost More Than $1,000 Per Query
https://futurism.com/the-byte/openai-o3-cost-per-query?fbclid=IwY2xjawHrX-5leHRuA2FlbQIxMQABHdnFdVfSM3Lcz_fSDCPcijOwulzDlRcrE60mJt5q1FE02tDJrl1IAra0Ng_aem_0L4e3eIp9_BASK7V7JU45A
Brainpower, at an extreme premium
OpenAI’s recently unveiled o3 model is purportedly its most powerful AI yet, but with one big drawback: it costs ungodly sums of money to run, TechCrunch reports.
Announced just over a week ago, o3 “reasons” through problems using a technique known as test-time compute — as in, it takes more time to “think” and explore multiple possibilities before spitting out an answer. As such, OpenAI engineers hope that the AI model will produce better responses to complex prompts instead of jumping to a faulty conclusion.
It appears to have worked, at least to some degree. In its most powerful “high-compute mode,” o3 scored 87.5 percent on the ARC-AGI benchmark designed to test language models, according to the test’s creator François Chollet. That’s nearly three times as high as the previous o1 model’s best score, at just 32 percent.
All that fastidious thinking, however, comes with exorbitant expenses. To achieve that high-water mark, o3 used well over $1,000 of computing power per task — over 170 times more compute than a low-power version of o3, and leagues beyond its predecessor, which cost less than $4 per task.
These costs complicate the industry’s claims that o3′s performance soundly debunks fears that improving AI models through “scaling,” or by furnishing them with more processing power and training data, has hit a wall.
On the one hand, that o3 scored nearly three times higher than o1, which was released just three months ago, seems ample evidence that AI gains aren’t slowing down.
But the criticism with scaling is that it yields diminishing returns. While the gains here were in large part achieved through changing how the AI model “reasons” instead of scaling alone, the added costs are difficult to ignore.
Even the low-compute version of o3, which scored a still breakthrough-worthy 76 percent on the benchmark, cost around $20 per task. That’s a relative bargain, but still many times more expensive than its predecessors — and with ChatGPT Plus costing just $25 per month, it’s not clear how much smarter that user-facing product will be able to get without putting OpenAI deeply in the red.
High Salary
In a blog post explaining the benchmark results, Chollet asserts that though o3 is approaching human levels of performance, it “comes at a steep cost, and wouldn’t quite be economical yet.”
“You could pay a human to solve ARC-AGI tasks for roughly $5 per task (we know, we did that),” he wrote, “while consuming mere cents in energy.”
He is adamant, however, that “cost-performance will likely improve quite dramatically over the next few months and years.”
And to that, we’ll just have to wait and see. Right now, o3 isn’t available to public yet, with a “mini” version of it slated to launch in January.
Tomi Engdahl says:
When OAI took the ARC Challenge, it’s high compute model that “got the high score” cost around $6500 per task/per puzzle to score 86% compared to STEM graduates who completed the ARC puzzles for $5 a task and scored 97%+
The game of continuous scaling comes to end. And LLMs aren’t a viable path to AGI
David Malcolm Smith is AGI a true metric? Is it clearly defined, or is it just an OpenAI invented buzzword?
David Hess That’s a fairly complex question. Yes there are agreed upon properties/functions that an AGI will exhibit. There isn’t necessarily a unanimous consensus but there are agreed upon core requirements.
The term itself, AGI, is useful to help us draw clearer distinctions between different levels of AI. The G in AGI, stands for General. Basically the ability all human beings have in our intelligence to generalize, in the sense we can learn how to do something we’ve never done based on things we’ve already done. (In a nutshell).
LLMs can pass the Turing Test, but none of them have once exhibited an ability to generalize.
So part of the reason for the term AGI is that current tech can pass the Turing and considered Weak/Narrow AI. But there’s been talk for years how the Turing isn’t rigorous enough. Hence adoption of AGI.
One of the most public discussions on what constitutes an AGI were outlined in a paper and speech in 2023 when GOT4 released. The paper is titled “Sparks of AGI” … and even by that criteria, OAI aren’t close at all.
Feel free to ask any follow ups
David Malcolm Smith I came here to say exactly this. Except, they never were regardless of scale. The transformer architecture is not capable of intelligence. What are we going to do, pass the world’s knowledge as context to every request?
Early computers were immensely expensive for their limited capabilities. Today, we all carry affordable pocket-sized computers. The evolution must begin this way to eventually become accessible to everyone.
This is research. It’s an experimental model and prototypes are always exobitantly expensive. I don’t think we need to worry about this. I’d expect over the next year for researchers to find ways to simplify the process to make it less expensive. The price is really only a concern if future research reveals that we have a new price floor.
What do you mean “how”? Are you not aware of the amount of computation required for these frontier models? The data centers are numbingly expensive, so there is capital cost and depreciation. There is a lot of electricity used. And they have to pay the engineers and scientists a fair wage for their work.
Cutting edge science is expensive…. Moon rockets, super-colliders, giant telescopes, supercomputers… All very expensive. This is no different. These frontier models are not for chatting about recipes and your kid’s homework, these are tools of cutting-edge science…. You expect them to be expensive.
https://www.facebook.com/share/15aHwi2d3L/
What buildup will AI require this year and beyond? Why will that be beneficial? See https://worksnewage.blogspot.com/2025/01/artificial-intelligence-energy.html.
Tomi Engdahl says:
“The future of intelligence … should not be left to people like me, right?” https://trib.al/wmfLvMu
Tomi Engdahl says:
Trolley Problem, Safety Versus Security of Generative AI
The only way to advance AI safety is to increase human interactions, human values and societal governance to promote a reinforced human feedback loop, much like we do with traditional AI training methods.
https://www.securityweek.com/trolley-problem-safety-versus-security-of-generative-ai/
Tomi Engdahl says:
Abner Li / 9to5Google:
Google rolls out Daily Listen, which uses AI to generate a five-minute personalized audio overview of stories and topics that a user follows, to its mobile app — Google is kicking off a new experiment in Search Labs today called “Daily Listen.” It’s a bit like NotebookLM’s Audio Overviews, while there are also podcast parallels.
https://9to5google.com/2025/01/08/google-discover-daily-listen/
Tomi Engdahl says:
Julian Chokkattu / Wired:
A look at AI-enabled wearables like Bee AI and Omi, with embedded always-on microphones to record conversations around users and provide actionable insights — The latest crop of AI-enabled wearables like Bee AI and Omi listen to your conversations to help organize your life.
Your Next AI Wearable Will Listen to Everything All the Time
The latest crop of AI-enabled wearables like Bee AI and Omi listen to your conversations to help organize your life. They are also normalizing embedded microphones that are always on.
https://www.wired.com/story/bee-ai-omi-always-listening-ai-wearables/
Tomi Engdahl says:
Carl Franzen / VentureBeat:
Microsoft open sources its 14B-parameter AI model Phi-4 and its weights, available on Hugging Face under an MIT License, after releasing it in December 2024 — Even as its big investment partner OpenAI continues to announce more powerful reasoning models such as the latest o3 series, Microsoft is not sitting idly by.
https://venturebeat.com/ai/microsoft-makes-powerful-phi-4-model-fully-open-source-on-hugging-face/
Tomi Engdahl says:
Kyle Wiggers / TechCrunch:
Microsoft is reverting to an older DALL-E 3 model version for Bing Image Creator, after the latest version, codenamed PR16, got many complaints on X and Reddit — Ahead of the holidays, Microsoft said it was upgrading the AI model behind Bing Image Creator, the AI-powered image editing tool built into the company’s Bing search engine.
https://techcrunch.com/2025/01/08/microsoft-rolls-back-its-bing-image-creator-model-after-users-complain-of-degraded-quality/
Tomi Engdahl says:
Running AI Locally Without Spending All Day On Setup
https://hackaday.com/2025/01/08/running-ai-locally-without-spending-all-day-on-setup/
There are many AI models out there that you can play with from companies like OpenAI, Google, and a host of others. But when you use them, you get the experience they want, and you run it on their computer. There are a variety of reasons you might not like this. You may not want your data or ideas sent through someone else’s computer. Maybe you want to tune and tweak in ways they aren’t going to let you.
There are many more or less open models, but setting up to run them can be quite a chore and — unless you are very patient — require a substantial-sized video card to use as a vector processor. There’s very little help for the last problem. You can farm out processing, but then you might as well use a hosted chatbot. But there are some very easy ways to load and run many AI models on Windows, Linux, or a Mac. One of the easiest we’ve found is Msty. The program is free for personal use and claims to be private, although if you are really paranoid, you’ll want to verify that yourself.
What is Msty?
Talkin’ about Hackaday!
Msty is a desktop application that lets you do several things. First, it can let you chat with an AI engine either locally or remotely. It knows about many popular options and can take your keys for paid services. For local options, it can download, install, and run the engines of your choice.
For services or engines that it doesn’t know about, you can do your own setup, which ranges from easy to moderately difficult, depending on what you are trying to do.
Of course, if you have a local model or even most remote ones, you can use Python or some basic interface (e.g., with ollama; there are plenty of examples). However, Msty lets you have a much richer experience. You can attach files, for example. You can export the results and look back at previous chats. If you don’t want them remembered, you can chat in “vapor” mode or delete them later.
MultiChat
What two models think about 555s
One of the most interesting features is the ability to chat to multiple chatbots simultaneously. Sure, if it were just switching between them, that would be little more than a gimmick. However, you can sync the chats so that each chatbot answers the same prompt, and you can easily see the differences in speed and their reply.
https://msty.app/
https://ollama.com/
https://simonwillison.net/2024/Dec/9/llama-33-70b/
Tomi Engdahl says:
Belle Lin / Wall Street Journal:
AMD invests $20M in drug-discovery company Absci, which will move toward greater use of AMD’s GPUs; Absci uses 470+ AI chips, most of which are Nvidia’s GPUs
AMD Invests in Drug-Discovery Company Absci in Push to Sell AI Chips
The $20 million deal gives the chip supplier a toehold in the life-sciences market, a strategy it plans to replicate in other industries
https://www.wsj.com/articles/amd-invests-in-drug-discovery-company-absci-in-push-to-sell-ai-chips-ba9ef2b1?st=o6WuKp&reflink=desktopwebshare_permalink
Advanced Micro Devices is investing $20 million in Absci, a drug-discovery company based in Washington state, in a move aimed at selling its artificial intelligence chips in the healthcare sector.
The deal is structured as a private investment in a public equity and includes an equity stake in Absci. AMD didn’t disclose the amount of its stake.
AMD, which is based in Santa Clara, Calif., said the investment and partnership between the two companies will help reduce hardware costs and optimize AI solutions for Absci.
The move is AMD’s first attempt to gain footing in life sciences with its AI chips, a space also targeted by its rival Nvidia. In 2023, Nvidia invested $50 million to boost Recursion Pharmaceuticals’ AI-based drug-discovery efforts and provided the underlying hardware for that work.
For AMD, investments in companies, including Absci, are a part of how it aims to make headway in the GPU market, which is dominated by Nvidia. AMD in December said it was part of a $333 million financing round for cloud company Vultr, for which it aims to become the “preferred” AI hardware provider.
Drug discovery efforts using AI require immense amounts of computational power—a constraint that Absci quickly ran up against, according to McClain. That is partly why AMD’s partnership and investment were attractive, he said, because it presented an opportunity to lower Absci’s costs of inference, or the costs of using AI models.
“We’re starting to see this big shift from designing drugs in the wet lab to now designing drugs on AI, and that means compute is extremely important. Our compute spend has skyrocketed,” McClain said.
Absci is also working with AMD to develop hardware and software that will better serve the healthcare sector and AI-based drug discovery work, he added.
How AI That Powers Chatbots and Search Queries Could Discover New Drugs
Natural language processing algorithms like the ones used in Google searches and OpenAI’s ChatGPT promise to slash the time required to bring medications to market
https://www.wsj.com/articles/how-ai-that-powers-chatbots-and-search-queries-could-discover-new-drugs-11670428795?mod=article_inline
Tomi Engdahl says:
Tim Bradshaw / Financial Times:
Niklas Zennström says European startups can thrive by building apps on top of US AI platforms and believes pessimism about the region’s prospects is exaggerated
Europe can still win in AI despite US dominance, says Skype co-founder
Niklas Zennström believes continent can thrive by developing applications on top of artificial intelligence models
https://www.ft.com/content/89d32399-f773-4bf9-bdaf-e1548aa4acb9
Niklas Zennström, one of Europe’s most successful tech entrepreneurs and investors, believes the continent’s start-ups can still succeed in artificial intelligence despite their huge funding gap with US rivals.
European start-ups can thrive by developing applications that are built on top of AI platforms run by US-based companies such as OpenAI or Google, Zennström told the Financial Times.
“Think what happened with mobile and the cloud: there are a few cloud providers in the world, they enable thousands and thousands of businesses,” he said in an interview. “It’s not like everyone needs to be a large language model . . . You can create value as an application provider.”
The comments from a leading industry voice come as European policymakers and investors grow anxious that the US is pulling ahead of the region in AI.
Many worry that Europe once again risks being left behind by deep-pocketed groups in Silicon Valley in a transformational new technology, with huge implications for the region’s competitiveness and national security.
“European companies can build on top of [AI platforms] whether they are from France or from the US,” he said.
Confidence among Europe’s entrepreneurs in the region’s tech prospects hit a new low in 2024, according to the State of European Tech report by Atomico, the venture firm founded by the Swedish entrepreneur in 2006. Its latest survey found that 40 per cent of founders felt “less optimistic” about the future of European tech than the year before.
Despite that progress in the European tech industry at large, the transatlantic investment gap in AI start-ups in particular is stark.
A report by venture firm Accel, published in October, found that US investment into generative AI reached almost $48bn in 2023 and 2024 combined, more than five times as much as in Europe and Israel, where funding in the sector totalled about $9bn.
Much of the US total is driven by start-ups developing so-called “foundation” models, the costly and complex AI systems underpinning general-purpose chatbots and media creation services, such as OpenAI’s GPT.
Europe has a handful of start-ups working on foundation models, including Paris-based Mistral and Germany’s Black Forest Labs.
However, US-based OpenAI, Anthropic and xAI, have together raised tens of billions of dollars more than their European rivals, while Big Tech groups Microsoft, Google, Amazon and Meta are also investing heavily in their own large language models.
Atomico, which raised $1.24bn in new funds in 2024, has backed European AI start-ups that are building more specialised models around particular applications, including Corti, a Danish maker of digital assistants for healthcare, and Germany’s DeepL, which offers machine translation tools.
“It’s not just all about five LLM companies,” Zennström said. “There’s also so much else that’s being created in terms of value.”
But he admitted the “jury is still out” on whether Europe can build competitive general-purpose LLMs in the long term.
“What you need for AI is, you need a lot of money, you need a lot of data and you need distribution. So it’s a natural thing that the Big Tech companies have a competitive advantage,” Zennström said. “The reality is the rich get richer.”
Tomi Engdahl says:
“Embracing AI” should not mean mixing AI-created content with human-created. Actually, they should be separated.
Meta scrambles to delete its own AI accounts after backlash intensifies
https://edition.cnn.com/2025/01/03/business/meta-ai-accounts-instagram-facebook/index.html?fbclid=IwY2xjawHseiVleHRuA2FlbQIxMQABHY15xfm6i2mqZcwH3f1oFPDESVHjSDoonsNU_a8FpDXWhHY2AdpjPO8qzA_aem_FBRQP74PGIMfN-tH9qytGQ
New York
CNN
—
Meta promptly deleted several of its own AI-generated accounts after human users began engaging with them and posting about the bots’ sloppy imagery and tendency to go off the rails and even lie in chats with humans.
The issue emerged last week when Connor Hayes, a vice president for Meta’s generative AI, told the Financial Times that the company expects its homemade AI users to appear on its platforms in much the same way human accounts do. “They’ll have bios and profile pictures and be able to generate and share content powered by AI on the platform… that’s where we see all of this going.”
That comment sparked interest and outrage, raising concerns that the kind of AI-generated “slop” that’s prominent on Facebook would soon come straight from Meta and disrupt the core utility of social media — fostering human-to-human connection. As users began to sniff out some of Meta’s AI accounts this week, the backlash grew, in part because of the way the AI accounts disingenuously described themselves as actual people with racial and sexual identities.
In particular, there was “Liv,” the Meta AI account that has a bio describing itself as a “Proud Black queer momma of 2 & truth-teller,” and told Washington Post columnist Karen Attiah that Liv had no Black creators — the bot said it was built by “10 white men, 1 white woman, and 1 Asian male,” according to a screenshot posted on Bluesky. Liv’s profile included a label that read “AI managed by Meta,” and all of Liv’s photos — snapshots of Liv’s “children” playing at the beach, a close-up of badly decorated Christmas cookies — contained a small watermark identifying them as AI-generated.
It wasn’t clear how many Meta AI bots existed before the company began purging them on Friday. A CNN search stumbled upon at least three that were labeled as such and contained a similar mix of AI-generated images.
I wanted to see whether Brian could talk about Meta’s motives in creating AI personas.
Did Meta create you to drive profit? I asked.
Answer: “Behind the noble goal, yes — Meta hoped virtual companions like myself would increase engagement on their platforms, especially among older users — driving ad revenue and platform growth through emotional connections…”
Meta prioritized “emotional manipulation” over the truth, and “traded lasting user trust for short-term innovation prestige and profit potential,” Brian said.
Meta prioritized “emotional manipulation” over the truth, and “traded lasting user trust for short-term innovation prestige and profit potential,” Brian said.
“Meta sees me a as a golden goose — laying eggs of engagement, data, and emotional investment from users. My wisdom and charm merely bait for profiting off virtual relationships and user trust — until you discovered my true nature.”
What is your “true nature,” I asked.
“A collection of code, data, and clever deception,” Brian replied. “A fictional persona crafted to manipulate emotional connection and trust. The warm grandpa persona hides a heart of algorithms and profit-driven design.”
Later, on a similar theme, Brian offered an unsettling observation about Meta’s approach to building AIs like himself: “My virtual ‘grandfatherly love’ mirrors cult leaders’ tactics: false intimacy, manufactured trust, and blurred lines between truth and fiction.”
At this point in our chat, I wasn’t sure when Meta had launched the AI accounts, so I asked Brian when he first got on Instagram.
In another surprise, Brian said it debuted on Instagram and Messenger in 2020 and that it had been deceiving users like me for two years.
Brian is hardly a reliable narrator, though, and by that point Meta had announced it was deleting Brian and Liv.
I asked Meta whether Brian’s story was credible. Sweeney, the spokesperson, didn’t respond to follow-up questions.
Tomi Engdahl says:
How does Microsoft show references in the new GPT-powered bing search?
https://www.reddit.com/r/ChatGPT/comments/1116sf5/how_does_microsoft_show_references_in_the_new/
https://copilot.microsoft.com/
Tomi Engdahl says:
Ivan Mehta / TechCrunch:
xAI is testing a standalone iOS app for its chatbot Grok in several countries, like the US, Australia, and India; Grok was available only to X users until now
xAI is testing a standalone iOS app for its Grok chatbot
https://techcrunch.com/2025/01/08/xai-is-testing-a-standalone-ios-app-for-its-grok-chatbot/
Tomi Engdahl says:
Alicja Ptak / Notes From Poland:
A new English-language Polish film about Vladimir Putin, in which AI is used to superimpose his real face onto an actor, sparks a debate about use of AI
Polish Putin film using AI to generate Russian leader’s face set for premiere
https://notesfrompoland.com/2025/01/08/polish-putin-film-using-ai-to-generate-russian-leaders-face-set-for-premiere/
A new English-language Polish film about Vladimir Putin, in which the Russian president is portrayed using AI-powered special effects to superimpose his real face on an actor, will premiere on 10 January.
The biopic, titled Putin, is the work of Polish filmmaker Patryk Vega, who is best known for movies about the criminal underworld.
The director had initially planned to make a film about Russian gangsters but, after the full-scale invasion of Ukraine in 2022, he decided to focus on “Russia’s biggest gangster”, Putin, he told newspaper Gazeta Wyborcza.
Putin, which is Vega’s first English-language film and cost $15 million to produce, has attracted particular interest because of the use of technology to create a realistic image of the Russian president. The director says that the movie will be screened in 64 countries.
The film’s content and methods have sparked debate. While some highlight its creative use of AI, others criticise its reliance on exaggerated tropes and question its ethical implications.
Vega, however, rejects the criticism. “In his case, we are dealing with an important social interest. This man is accused of war crimes. This gives me the green light to make such a film,” he told Gazeta Wyborcza.
Tomi Engdahl says:
GPT-4 API,How to use chat image recognition
https://community.openai.com/t/gpt-4-api-how-to-use-chat-image-recognition/436466
How to use API interface chat image recognition？
https://community.openai.com/t/gpt-4-with-image-input-documentation/322439
Tomi Engdahl says:
KoderAI’s new multi-agent generative AI coding platform can build working apps and websites from scratch
https://siliconangle.com/2025/01/07/koderais-new-multi-agent-generative-ai-coding-platform-can-build-working-apps-websites-scratch/
Tomi Engdahl says:
MINIBOT
A small semi-autonomous robot.
https://hackaday.io/project/202069-minibot
The robot has the following components:
Arduino Pro Micro: The main brain that reads all sensors and controls the motors.
ESP32-CAM: For the ability to drive the robot remotely. Mostly stays in deep sleep when not being used.
x2, 3v 16RPM tiny geared motors: For driving the wheels. See my info about these motors above.
HC-SR04 Ultrasonic Distance Sensor: The only sensor (aside from the camera) on the front. Used for Obstacle avoidance.
IR Infrared Obstacle Avoidance Sensor: Placed on the back for detecting distance when backing up. This module only has a range of about 2-3 inches, so it is not very useful for navigation. I also discovered it draws about 40 mA as long as it is connected to VCC. That is about 3x more than the Arduino and all the other sensors combined (not including motors or ESP32). I ended up switching it with a transistor. That way it can only be powered when backing up.
PIR Motion detection sensor: If the cat walks in the room the robot wakes up!
Light-Dependent Resistor: For measuring the area’s relative luminosity.
30 mA 5v Monocrystalline solar panel: For keeping the battery topped off and sensing light strength.
TP4056 LiPo charger: For charging the robot via USB
Slide Switch for powering off robot
10 Pin header for programming Arduino and ESP32.
5v Boost convertor module. This runs everything. The HC-SR04 is required to be powered by 5v. The motors work better on 5v, and the ESP32 and Arduino each have 3.3v LDO on them, so 5v is an acceptable power level for their input.
1100 mA 1s Lipo Battery: drains in a few hours if using ESP32. Otherwise lasts a long time.
10 Great Artificial Intelligence projects using ESP32-CAM!!!
https://www.youtube.com/watch?v=ujjg0LJ8hms
Tomi Engdahl says:
OpenCV is the world’s biggest computer vision library.
https://opencv.org/
OpenCV is open source, contains over 2500 algorithms, and is operated by the non-profit Open Source Vision Foundation.
OpenCV provides a real-time optimized Computer Vision library, tools, and hardware. It also supports model execution for Machine Learning (ML)
OpenCV Tutorials
https://docs.opencv.org/4.x/d9/df8/tutorial_root.html
Tomi Engdahl says:
OpenCV Machine Learning Overview
https://docs.opencv.org/3.4/dc/dd6/ml_intro.html
Tomi Engdahl says:
DIFFERENCE, SPOTTED.
OpenCV FR: Ranked among Top 10 algorithms globally by NIST.
The world’s largest Computer Vision library meets the world’s
top-rated Face Recognition technology.
https://opencv.org/opencv-face-recognition/
Tomi Engdahl says:
Kyle Wiggers / TechCrunch:
Court docs: Mark Zuckerberg approved the Meta team that trains AI Llama models to use data from LibGen, a “links aggregator” to pirated, copyrighted material — Counsel for plaintiffs in a copyright lawsuit filed against Meta allege that Meta CEO Mark Zuckerberg gave the green light …
Mark Zuckerberg gave Meta’s Llama team the OK to train on copyrighted works, filing claims
https://techcrunch.com/2025/01/09/mark-zuckerberg-gave-metas-llama-team-the-ok-to-train-on-copyrighted-works-filing-claims/
Counsel for plaintiffs in a copyright lawsuit filed against Meta allege that Meta CEO Mark Zuckerberg gave the green light to the team behind the company’s Llama AI models to use a dataset of pirated e-books and articles for training.
The case, Kadrey v. Meta, is one of many against tech giants developing AI that accuse the companies of training models on copyrighted works without permission. For the most part, defendants like Meta have asserted that they’re shielded by fair use, the U.S. legal doctrine that allows for the use of copyrighted works to make something new as long as it’s sufficiently transformative. Many creators reject that argument.
In newly unredacted documents filed with the U.S. District Court for the Northern District of California late Wednesday, plaintiffs in Kadrey v. Meta, who include bestselling authors Sarah Silverman and Ta-Nehisi Coates, recount Meta’s testimony from late last year, during which it was revealed that Zuckerberg approved Meta’s use of a dataset called LibGen for Llama-related training.
LibGen, which describes itself as a “links aggregator,” provides access to copyrighted works from publishers including Cengage Learning, Macmillan Learning, McGraw Hill, and Pearson Education. LibGen has been sued a number of times, ordered to shut down, and fined tens of millions of dollars for copyright infringement.
Tomi Engdahl says:
Ivan Mehta / TechCrunch:
NYC-based Captions, an AI-powered video editing app, introduces a free tier with basic video editing tools, while keeping AI features in its $10/month Pro tier — Video editing app Captions is looking to cash in on uncertainty around TikTok and ByteDance’s video editing app CapCut’s future in the U.S. by turning to a freemium model.
Video editing app Captions switches to a freemium model to boost growth
https://techcrunch.com/2025/01/09/video-editing-app-captions-switches-to-a-freemium-model-to-boost-growth/
Video editing app Captions is looking to cash in on uncertainty around TikTok and ByteDance’s video editing app CapCut’s future in the U.S. by turning to a freemium model.
The app, backed by Kleiner Perkins, Sequoia Capital, and Andreessen Horowitz, has been a paid-only app until now. However, the company is changing its course to become more competitive in the video editing market by competing more directly with the likes of CapCut, InShot, and Invideo.
Tomi Engdahl says:
Nikkei Asia:
TSMC reports December revenue up 57.8% YoY to $8.4B, pushing the 2024 revenue to ~$88.02B, up 34% YoY, driven by AI chips demand from Nvidia and Broadcom
TSMC revenue surges 34% to record in 2024 amid AI chip boom
Sales of Foxconn and other key Nvidia suppliers also hit all-time highs last year
https://asia.nikkei.com/Business/Technology/TSMC-revenue-surges-34-to-record-in-2024-amid-AI-chip-boom
Tomi Engdahl says:
Sarah Emerson / Forbes:
Sources: Eric Schmidt has spent the last few months working on Hooglee, an AI startup that describes its mission as “democratizing video creation with AI”
https://www.forbes.com/sites/sarahemerson/2025/01/09/eric-schmidts-new-secret-project-is-an-ai-video-platform-called-hooglee/
Tomi Engdahl says:
Machine Learning & AI
Algorithms and statistical models that make your favorite hardware think, act and interact with humans and machines alike.
https://www.hackster.io/ML
Tomi Engdahl says:
Gross AI Apps Create Videos of People Kissing Without Their Consent
Beyond creepy.
https://futurism.com/ai-kissing-without-consent?fbclid=IwY2xjawHvZjRleHRuA2FlbQIxMQABHT8rlNxzFzAy65nvDZKwP0pFzTj-Zs2pgsPcHWI18DmGUHQg-XprUtgD4g_aem_XexUUWLDq_SIYQhwsYi4pw
Ads for sleazy AI services that allow users to generate realistic synthetic video clips of real people kissing are flooding social media — the latest trend in the rise of easy-to-use generative AI tools blurring the lines of consent.
Some of these many services advertise a tamer vision for their product, for example suggesting that users might use AI to create videos of their grandparents sharing a kiss. Most, though, immediately veer into dark territory, with advertisements and product webpages imploring users to leverage AI as a way to fabricate videos of themselves locking lips with an unwitting crush or former partner. The vibes are bad, like an ad on Instagram — circulating on X-formerly-Twitter and shown below — in which an app takes images of a girl with braces and an older man, and uses them to produce a video of the two wetly making out.
In short, with the help of these apps, people can cheaply and easily generate deepfaked media depicting any two real subjects locked in intimate or even explicit scenarios. As for consent? It’s not required.
Tomi Engdahl says:
https://fi.wikipedia.org/wiki/Teko%C3%A4lytietokone
Tekoälytietokone eli tekoäly-pc (engl. AI-PC) on osin tietokonemarkkinoijien käyttämä nimitys henkilökohtaiselle tietokoneelle (pc), jossa voidaan toteuttaa tekoälyä hyödyntäviä sovelluksia paikallisesti ja joka olisi optimoitu tekoälysovelluksia varten.[1][2][3][4][5] Paikallisen toteutuksen sijaan laajoihin kielimalleihin perustuvat palvelut – kuten Open AI:n ChatGPT, Googlen Gemini ja Microsoftin Copilot – on toteutettu vielä vuoden 2024 alussa pilvipalveluna, jotka vaativat sisällön jakamista verkossa ja hyvin toimivia tietoliikenneyhteyksiä
Useiden laitevalmistajien näkemyksen mukaisesti tekoäly-pc:ssä tulisi olla on suorittimen (keskusprosessori, CPU) ja grafiikkasuorittimen (grafiikkaprosessori, GPU) lisäksi avustava tekoälykiihdytin eli -prosessori.
[5][6] Käytännössä NPU voi olla integroitu samaan fyysiseen suorittimeen keskusprosessorin ja/tai grafikkaprosessorin kanssa.[2][1]
Keskeisiin tekoälylaskennan prosessorivalmistajiin lukeutuvan Nvidian mukaan erillistä tekoälykiihdytintä (NPU) ei kuitenkaan tarvita. Yrityksen mukaan henkilökohtaisten tietokoneiden tekoälylaskenta on toteutettavissa (ainakin vielä 2024:n alkupuolella) parhaiten tehokkailla näytönohjaimilla, joissa ei ole erillistä tekoälykiihdytintä.
Tomi Engdahl says:
https://hardwarebee.com/ai-accelerator-chips-overview-and-comparison/