AI Archives - Page 5 of 9

Anthropic: Virtual employees will arrive next year

Virtual employees are step uup mere agents, say Anthropic. — Virtual employees are a step up mere agents and could be roaming the offces pretty soon. (Picture: Anthropic)

In a recent interview with Axios focusing mostly on security issues, Anthropic said «virtual employees» will be a step up from using mere «agents» on corporate networks.

This will be the next AI innovation, said Jason Clinton, the company’s chief of information security.

Whereas agents can focus on specific, programmable tasks, acts with some autonomy and of course require oversight, a «virtual employee» takes it a step further, with having their own memories and their own corporate accounts and passwords.

This is a major headache for cybersecurity, Clinton further explained, about oversight and hackability of the new employees.

Go read the full story here – at Axios, and check out Anthropics research on agents.

Cluely AI app will let you «cheat on anything»

Cluely AI is a "hidden" window over your browser that gives you the answers you need. — Cluely AI is a “hidden” window over your browser that gives you the answers you need. (Picture: Cluely)

The new AI startup is founded by Chungin “Roy” Lee and Neel Shanmugam, who recently made waves for using AI to cheat at Columbia University and at Amazon. Now they are going pro with a new app, hidden from even screen recorders.

AI cheating is, of course, a huge problem at schools and colleges and even for remote job interviews already, with many returning to in-person, written exams, and now these two are taking it to the next level.

Continue reading “Cluely AI app will let you «cheat on anything»”

Netflix CEO says AI can make movies better, not just cheaper

AI puts advanced tools in the hands of low budget creators, says Co-CEO Ted Sarandos. (Picture: Netflix)

James Cameron says we need AI in blockbuster movies to make them 50% cheaper. Netflix Co-CEO Ted Sarandos rebuts this week that we can use it to make them «10% better» instead.

The debate started last week when Cameron took to a podcast to explain his board role at Stability AI, claiming he wanted to «figure out» the technology and how it can help in movies.

— If we want to continue to see the kinds of movies that I’ve always loved and that I like to make and that I will go to see — ‘Dune,’ ‘Dune: Part Two,’ or one of my films or big effects-heavy, CG-heavy films — we’ve got to figure out how to cut the cost of that in half, Cameron said, according to Variety last week.

Continue reading “Netflix CEO says AI can make movies better, not just cheaper”

ChatGPT o3 and o4-mini are big steps toward AI agents

The models are inching ahead in benchmarks, but multimodality is where they truly shine. (Picture: OpenAI)

OpenAI’s latest model drop hints at a future where agents can do most of our work — and is proving the point with image processing.

The new reasoning models are managing an ever so slight lead in many benchmarks and therefore earns the right to be called state of the art, but of particular note is that they improve on GPT o1 and o3-mini by almost 30% in the coding benchmark SWE-Bench Verified, OpenAI claims in their launch post.

— These are the smartest models we’ve released to date, representing a step change in ChatGPT’s capabilities for everyone from curious users to advanced researchers, says OpenAI.

Continue reading “ChatGPT o3 and o4-mini are big steps toward AI agents”

New ChatGPT 4.1 coding family drastically reduces costs

The GhatGTP 4.1 are smarter than 4o, but are otherwise bang on average. (Picture: OpenAI9

Sam Altman of OpenAI teased this weekend that we were in for a week of big launches, and chatter was about an open source model that would be better than anything available.

On Monday night, though, the AI company launched ChatGPT 4.1, 4.1-mini and 4.1-nano, which at first seem a bit underwhelming.

While you might be getting tired of ChatGPT’s word salad lineup, these latest models offer mostly cost efficiency and larger input levels for coding tasks. They won’t be available in the app — but are instead rolled out on the API.

Continue reading “New ChatGPT 4.1 coding family drastically reduces costs”

ChatGPT memory now includes all past chats

No more re-entering facts in ChatGPT. It now remembers the past. (Picture: OpenAI)

In a huge move for ChatGPT, all previous chats will be considered something akin to the «memory»-feature in coming updates to the Plus and Pro segments.

The feature is supposed to make ChatGPT more personal and relevant to users, and you wont have to manage the paltry «memories»-storage any more, writes TechCrunch.

Here’s the official announcement from OpenAI:

Continue reading “ChatGPT memory now includes all past chats”

— This is impressive research! John Carmack defends Microsoft’s AI Quake

Industry legends are coming in force to support AI tools. (Picture: id Software)

The legendary co-creator of the Doom and Quake games for id Software, came out swinging on Monday in support of AI game tools, citing how the industry has developed from the early days of hand coding hex digits.

The heart of the issue is the release of Microsoft’s Copilot Labs’ release of an AI powered, fully playable tech demo of new Quake II maps this weekend.

It wasn’t made by hand or by coding, but rather built as an ai world you could navigate freely, and the frames are rendered on the fly — at 10 frames per second with a a resolution of 640 x 360.

Continue reading “— This is impressive research! John Carmack defends Microsoft’s AI Quake”

Meta gamed the LMArena AI benchmarks with new model

People are astounded Meta used a non public, unreleased and optimized model on the industry’s most respected benchmark. (Picture: Meta)

According to several reports, it seems Meta used an unpublished Llama 4 Maverick model created especially to score well on the LMArena benchmark.

Surprisingly good ranking
The largest selling point for their latest Maverick model was how well it did in precisely this benchmark, scoring just above ChatGPT 4o-latest and and slightly below Gemini 2.5 Pro, considered the current cutting edge of AI engineering.

The fact that Llama 4 Maverick got second place in between these, on the most watched leaderboard in AI, raised a lot of eyebrows over the weekend.

Continue reading “Meta gamed the LMArena AI benchmarks with new model”

Meta drops Llama 4 models, with «mixture of experts»

The Llama 4 models take off to a great start with the benchmarks, runs on little hardware, and is really cost efficient. — The Llama 4 models seem tailor made for STEM benchmarks, runs on little hardware, and is very cost efficient. (Picture: LadyDragonflyCC, CC BY 2.0)

The latest models out of Meta, the Llama 4 Maverick and the Scout, aren’t reasoning models, but put in extra work by using a panel of «experts» instead.

The models are rolling out right now on WhatsApp, Messenger and Instagram Direct, where they are free for everyone with a membership.

Will answer «contentious» questions
According to TechCrunch, the new models will answer questions previous models wouldn’t, responding to «debated» questions and «contentious» prompts «without judgement:»

— We’re continuing to make Llama more responsive so that it answers more questions, can respond to a variety of different viewpoints, and doesn’t favor some views over others, a meta spokesperson told the website.

Continue reading “Meta drops Llama 4 models, with «mixture of experts»”

Copilot labs show off new, playable Quake II levels

A picture of the AI rendered, playable Quake II. — Miss Quake II? Now you can play new levels generated by AI. (Picture: Screenshot)

Amid Ai fears in the gaming industry, replete with layoffs and questions about artistry in the age of AI, Microsofts Copilot labs has offered a glimpse into what the future of gaming might look like.

They spent only just over a week on training to produce a real time game model of Quake II, says Tom’s Hardware, that now plays a somewhat choppy 10 frames per second or so version of the game — with new, never before seen gameplay, generated on the fly.

Continue reading “Copilot labs show off new, playable Quake II levels”

UN: AI market to reach $4.8 trillion, 40% of jobs to be affected

The United Nations are worried developing nations will miss out on AI. (Picture via www.localdigital.com.au)

The UN Trade and development organization doesn’t think AI will simply go away, predicting instead that nearly half of all jobs will be affected by 2033.

They also warn that only a hundred corporations control the technology, and say the difference between developed and developing nations will grow as the technology gets unevenly rolled out.

Continue reading “UN: AI market to reach $4.8 trillion, 40% of jobs to be affected”

OpenAI raises $40 Billion, teases reasoning open weights model

OpenAi is now a $300B company, and is pushing into new frontiers. (Picture: OpenAI)

In a blockbuster day for OpenAI news, it also says it added 1 million users an hour last night, and is opening the famous image generator for free users.

The money raised by the ChatGPT maker values the company at $300 billion, and is the most money raised by a private tech company ever, as it was the last time they raised a mere $6.6 billion back in October 2024.

In fact, the funding is three times the previous high note in funding for tech companies.

Continue reading “OpenAI raises $40 Billion, teases reasoning open weights model”

Gemini 2.5 Pro now available for free, with strict qouta

Gemini 2.5 is topping leaderborads all over. (Picture: Google)

The latest Gemini launch got overshadowed by ChatGPTs new image generation, but the model has been absolutely crushing it in the benchmarks and averages.

The Pro 2.5 is currently one of the best large language model out there, and it is now available for free – with no set qoutas or limits on prompts.

Continue reading “Gemini 2.5 Pro now available for free, with strict qouta”

Claude AI reveals surprising internal thinking, says Anthropic

The internal thinking of an AI has been mostly opaque before this study. (Image: Anthropic.)

Some people think large AI models are preprogrammed by lots of people, whereas the opposite is often true.

Training an LLM like Claude often consists of unmonitored consumption of huge amounts of data, with minimal human involvement.

— Language models like Claude aren’t programmed directly by humans, says Anthropic, — They arrive inscrutable to us, the model’s developers. This means that we don’t understand how models do most of the things they do, they add.

Tracing the thoughts of an LLM
Now they have set out to change that, with a couple of scientific studies mapping out the internal reasoning, or how the model actually thinks in response to normal prompts.

Continue reading “Claude AI reveals surprising internal thinking, says Anthropic”

Anthropic scores early copyright win in battle with music publishers

A US court has denied a coalition of record companies’, including Universal Music, request to keep music lyrics out of the training data for their AI, finding that no immediate harm had been proved.

This is just one of many lawsuits against AI companies claiming they copied and used copyrighted materials in training their models, that hinges on the fair use provision of the copyright laws

This resurfaced overnight after Chatgpt’s new image generator was used en masse to produce images in the style of Studio Ghibli, which had obviously been used in training the model in so far that it could easily mimic their style.

Continue reading “Anthropic scores early copyright win in battle with music publishers”