Artificial Intelligence - Page 27

Chinese AI company DeepSeek is reportedly working on a new "semiconductor" venture, a new in-house AI chip that would compete directly with the likes of NVIDIA, AMD, and other companies.

In a new report from DigiTimes, we're learning that DeepSeek is looking into making its own AI chips, adding that the company has started a "major recruitment drive" hiring semiconductor experts to lead the AI chip project. Chinese companies have multiple hurdles in their way, the biggest being securing the semiconductor equipment and processes because of global sanctions.

The only chip access the likes of DeepSeek has access to would be SMIC, but they're many years behind the rest of the semiconductor industry, namely TSMC.

Meta recently announced two new breakthroughs from their global research labs. The first highlights that Meta successfully developed a model that can successfully convert brain signals into text. The second reveals how the brain transforms thoughts into words, offering new insights into language processing and AI development.

The first paper refers to an AI model that can decode up to 80% of typed sentences solely from brain activity. The model utilizes noninvasive MEG and EEG recordings, essentially sparing the need to conduct surgical procedures. It presents a few potential benefits, for example, the ability to establish brain-computer interfaces for those who have lost speech and clinical use for brain injury patients.

While the advancement is impressive, researchers are still encountering challenges with accuracy. While the AI can decode up to 80% of characters, errors still occur, making full, reliable communication difficult. It also requires stillness and a magnetically shielded room, which poses issues for practicality. So don't worry - you won't need to build a Magneto-style thought shielding helmet any time soon. It's also unclear how well this would work for patients with brain injuries and disorders.

Tiktok's parent company, ByteDance, has released another AI text-to-video model. The project, entitled Goku, allows users to convert text into hyper-realistic ad videos, many of which draw clear inspiration from the real-life reels we see on social media.

The project's GitHub page showcases examples of the tool in action, ranging from landscape demos to portrait videos to marketing footage. Some of the most jarring examples are the ones that look the simplest: people eating food, unboxing products, or doing their makeup. While you can still notice that uncanny AI feeling to many of them, the gap is beginning to close on being able to differentiate AI content from real ones.

The Goku model is a 'flow-based video generative foundation model' jointly developed by the University of Hong Kong and ByteDance. Released in February 2025, industry insiders believe that the newest model 'significantly improved' the visuals' consistency, controllability, and richness. The model is based on the 'rectified flow transformer architecture' and can generate images and videos with simple text prompts.

Samsung is reportedly working on redesigning its 6th-generation 1c DRAM to improve yield rates, in order to catch up

with its next-generation HBM4 process.

Samsung considering tweaking its 1c DRAM process is a key factor in the success of its next-gen HBM4, with a new piece from ZDNet Korea reporting that the South Korean memory giant is looking at designs for its cutting-edge DRAM processes, something it has been doing since 2H 2024, and that the company has redesigned its high-end 1c DRAM to sure its next-gen HBM processes are adopted into the AI industry, unlike its HBM3 memory being denied by NVIDIA.

US Vice President JD Vance gave a fantastic speech at the AI Action Summit in Paris on Tuesday, where he said the Trump administration "will sure that the most powerful AI systems are built in the US with American design and manufactured chips".

President Trump has been steadfast in wanting the USA to benefit from all of its silicon prowess, and now his running mate VP Vance has said the same thing.

VP Vance said: "The Trump administration believes AI will have countless revolutionary applications in economic innovation, job creation, national security, healthcare, free expression and beyond. To restrict its development now will not only unfairly benefit incumbents in the space, it would mean paralyzing one of the most promising technologies we have seen in generations".

Former Google CEO Eric Schmidt spoke to the BBC to discuss his concerns about artificial intelligence, highlighting a less commonly expressed fear: the possibility that AI could be weaponized for terror.

Schmidt, who led Google from 2001 to 2017, has been a strong advocate for responsible AI development. While the company is pushing towards artificial general intelligence (AGI) with heavy investments in Google Gemini, Schmidt has remained insistent on prioritizing safety throughout the process:

"It's really important that governments understand what we're doing and keep their eye on us."

According to a recent report from Colombia One, researchers are banding together to create an AI that facilitates animal-to-human translation for the first time in history. The Coller-Dolittle Prize has offered a $500,000 reward, or $10 million in equity investments, to the researchers who make the most progress in this field.

There are currently two leading projects in the race to talk to animals. Project Ceti (Cetacean Translation Initiative), which focuses on decoding whale communication using AI. And the Earth Species Project, which aims at broader translations of animal communication, including work on crow communication.

Project Ceti's work on whales involves using AI to translate the 'click' communication used by humpback and sperm whales. The project has been in the works for over a decade and has reportedly made 'some progress' towards the ambitious goal. However, researchers are running into challenges with the lack of available data. Project Citi is currently working with just 8,000 vocalization samples, which limits the extent to which they can develop a working mode.

In a recent X post, Sam Altman announced that OpenAI has decided to cancel the release of the full o3 model, which was scheduled to launch in February or March.

He cited the 'complexity' of their model and product offerings as a contributing factor and that they wished to simplify their product range. As part of the new strategy, OpenAI will be focusing all efforts towards the release of GPT-5. Which will incorporate elements of o3, along with its own features.

Last month, Sam Altman also cautioned AI fans to temper their expectations surrounding the company's progress towards superintelligence. He stated in his personal blog, "Twitter hype is out of control again," and that OpenAI will not be deploying AGI next month. He went on to mention that OpenAI is building some 'cool stuff', but for fans to cut their expectations 'by 100x'.

AI won't take your job, but it will reshape it in ways we've never seen before.

In a recent interview with World Wide Technology, NVIDIA CEO Jensen Huang shared his vision for AI's impact on the workforce, explaining that while no AI can fully replace a human job, many can automate up to 80% of what we do.

OpenAI CEO Sam Altman recently stated in an interview he feels sorry for Elon Musk's lifelong "insecurities" after OpenAI rejected Musk's bid to purchase the company for $97.4 billion.

The feud between Musk and Altman has certainly been heating up as of late and then reached a boiling point on Monday when the Wall Street Journal reported Musk offered to purchase OpenAI for a whopping $97.4 billion. However, that bid was rejected by OpenAI, and the company's CEO, Sam Altman, took to X to post a rebuttal offer for X, writing, "No thank you but we will buy Twitter for $9.74 billion if you want." In response to that X post from Altman, Musk wrote, "Swindler."

But now things have seemingly taken a more personal turn, as Altman has stated during a recent Bloomberg interview at the Paris AI Action Summit that this recent offer from Musk is just one of the many tactics the Tesla CEO has attempted in the past. When asked what Musk's goal would be with this offer Altman said he believes Musk is trying to slow OpenAI down due to Musk's AI company, xAI, being a direct competitor. However, Altman said he wishes Musk would compete by "building a better product," but "I think there's been a lot of tactics. Many, many lawsuits, all sorts of other crazy stuff. Now this."

OpenAI is fishing for money right now, with Elon Musk offering to buy the ChatGPT creator for $97.4 billion but being turned down by CEO Sam Altman... and now Epic Games boss Tim Sweeney wants to buy OpenAI... with VBucks, the virtual currency from Fortnite.

In a post on X, the founder and CEO of Epic Games said: "Will OpenAI take VBucks in payment? Thinking of making an offer" in reply to a post about a group of investors led by Elon Musk offering a $97.4 billion bid to takeover OpenAI. OpenAI CEO Sam Altman declined the $97.4B offer, with a counter-offer to acquire X (which he called Twitter) for $9.74 billion.

Musk said in a statement provided by his lawyer Mark Toberoff to The Wall Street Journal: "It's time for OpenAI to return to the open-source, safety-focused force for good it once was. We will make sure that happens". Better yet, Musk's team has promised to match or exceed any higher offers for OpenAI that it receives, and with the likes of Microsoft and other AI giants swirling around, things are going to get real interesting for OpenAI, real soon.

SpaceX, Tesla, and xAI boss Elon Musk has offered to buy ChatGPT maker OpenAI for $97.4 billion, to which CEO Sam Altman declined by saying "no thank you" while offering a $9.74 billion offer for X (which he trolls, and calls Twitter).

OpenAI is currently converting into a for-profit structure, away from its non-profit structure now, raising $40 billion at a $340 billion valuation, simultaneously launching a $500 billion AI infrastructure project. Elon Musk co-founded OpenAI with Altman back in 2015 before leaving in 2019, and now the Tesla and DOGE boss is offering less than one-third of what OpenAI's latest valuation was to acquire the company.

Musk said in a statement provided by his lawyer Mark Toberoff to The Wall Street Journal: "It's time for OpenAI to return to the open-source, safety-focused force for good it once was. We will make sure that happens". Better yet, Musk's team has promised to match or exceed any higher offers for OpenAI that it receives, and with the likes of Microsoft and other AI giants swirling around, things are going to get real interesting for OpenAI, real soon.

At the Artificial Intelligence Action Summit in Paris, the Trump administration refused to sign a global declaration that aims to encourage the "open," "inclusive," and "ethical" development of AI.

Much of the summit's discussions centered on AI regulation, with French President Emmanuel Macron stressing the need for clear ground rules to ensure the safe advancement of the technology.

"We need these rules for AI to move forward," Macron stated.

Researchers from Fudan University used Large Language Models (LLMs), the underpinning technology of AI, to determine if AI was capable of cloning itself.

The team pitted LLMs from Meta and Alibaba across 10 trials and found that both of the AI models created a clone of themselves in 50% of tests (Meta's Llama31-70B-Instruct model) and 90% of tests (Alibaba's Qwen2.5-72B-Instruct model.) The results, which are yet to be peer-reviewed, indicate each of the AI models may already have the capabilities of cloning itself outside the confinements set by humans, commonly referred to as "gone rogue."

The study was conducted with off-the-shelf GPUs and in a precisely controlled environment, with the researchers attempting to recreate a real-world environment. The team looked at two specific scenario: "shutdown avoidance" and "chain of replication". In the first scenario, the team programmed the AI models to detect when it was going about to be shut down and then clone itself before the termination process took place. The other scenario involved the AI's being programmed to clone itself and then program its cloned version with instructions to replicate itself. Creating a domino effect of cloned AI replicas.

Researchers at Chinese AI company ByteDance have unveiled their latest AI video model: OmniHuman. This end-to-end AI framework can generate realistic human videos from a single image, producing life-like movements and natural gestures while seamlessly integrating different audio and video inputs.

Several examples of OmniHuman in action have surfaced on X, featuring figures like Taylor Swift, Albert Einstein, and NVIDIA CEO Jensen Huang. Notable demonstrations include:

OmniHuman's capabilities are particularly striking. The Albert Einstein video, for example, was recreated from a still portrait taken in 1921. You can note slight shimmers in the recreation of the famous physicist, but the output is very convincing. From the animation of his hands and eyes, to the lip-syncing of the image to the audio provided.

Andrew Ng, the founder of DeepLearning AI, released a video to his socials last week outlining the release of Agentic Object Detection. While on the surface, this feature allows users to identify images using text prompts, there's a few additional layers that make it a significant advancement in computer vision.

As outlined in the company's video demonstration, it works by filling in a text prompt for the object you want to identify. In the example, he asks the tool to locate 'unripe strawberries' in an image that contains a mixture of both. As you can see in the image below, the tool successfully places a box around any visual element that fits the criteria.

So great - we can accurately locate different varieties of fruit. What makes this unique? To explain that, we'll have to go back to the training process.

A recent report from researchers at Microsoft and Carnegie Mellon suggests that generative AI use may contribute to a decline in critical thinking. The logic is simple: the more you rely on AI, the less critical thinking you engage in. However, the study-conducted on 319 knowledge workers-takes a deeper look at this issue, examining its real-world implications.

The conclusions were based on self-reported survey responses, where participants detailed their AI usage, how often they applied critical thinking, and the level of effort required.

One of the findings was that excessive reliance on AI can weaken cognitive abilities. Researchers found that respondents often mistook "copy-pasting with minor tweaks" for critical thinking-accepting AI-generated content without questioning its accuracy. This is particularly concerning given AI's tendency to be confidently wrong. Just look at some of Google's AI-generated summaries and their hilarious misfires as an example.

NVIDIA is currently the only supplier of high-end AI hardware, but that is about to change as OpenAI is reportedly nearing completion of its first generation, in-house AI-dedicated silicon.

The news comes from an exclusive report from Reuters that states the ChatGPT maker is currently in the midst of finalizing the design for its first in-house AI chip, and over the next few months, the design will be locked and sent over to the Taiwan Semiconductor Manufacturing Co (TSMC) for fabrication. OpenAI previously set a goal of hitting mass production of its own AI hardware sometime in 2026, and according to the report, these plans coincide with that timeline and illustrate the ChatGPT marker is on track for hitting them.

Despite OpenAI nearing the design completion for the chip, the challenges don't stop there as the first chip design will take approximately 6 months to produce, cost tens of millions of dollars, and there's no guarantee the silicon will function as intended. A failure in the first generation of silicon will mean OpenAI will need to take the design back to the drawing board, assess where it went wrong, and repeat the whole process all over again.

The Paris AI Action Summit is officially underway. All eyes are on the US and China, as discussions begin regarding the future of work, AI governance, AI for public good, and the risks of AI.

A key talking point in the lead-up to the event surrounds the US administration's stance on AI. President Trump has been vocal about his desire to make the US the 'world capital of artificial intelligence', quickly implementing his administration's own AI policies shortly after being sworn in.

"[We] must develop AI systems that are free from ideological bias or engineered social agendas," he said.

For sophisticated AI chatbots to exist, they need to be trained on large swaths of data, but where things get murky is when the big question is posed to the companies behind these AI chatbots - where did you get this data from? And was it obtained legally?

Since their massive rise in popularity, companies behind these AI chatbots have been accused of stealing copyrighted data, which is then used by the AI for training purposes to further increase its sophistication and, ultimately, the price the company charges for access to the AI. Obtaining datasets legally means companies must pay a licensing fee for copyrighted material, and also agree to a bunch of hoops set by the owner of the data. Why go through all that expenditure and stipulations when the dataset can just be pirated in the same way a movie can be illegally downloaded?

Companies such as OpenAI are currently embroiled in copyright lawsuits for these reasons, but OpenAI isn't the only AI company facing copyright lawsuits, as a lawsuit against Meta has recently been leaked online that accuses the Mark Zuckerberg-run company of obtaining 82 terabytes (TB) of books from an illegal source for AI training. The lawsuit states Meta illegally downloaded the contents of the books from "shadow libraries" such as Anna's Archive, Z-Library, and LibGen, with the suit quoting a Meta researcher who was against the use of pirated material.