Feed aggregator

Louis DeJoy said he plans to step down after a nearly five-year tenure marked by the coronavirus pandemic, surges in mail-in election ballots and efforts to stem losses through cost and service cuts.

(Image credit: Jose Luis Magana)

The reasons for the beachings are unclear. Reasons could include disorientation caused by loud noises, illness, old age, injury, fleeing predators and severe weather.

(Image credit: AP)

Despite promises to improve the system, problems within prisons keep mounting. Critics say the Dominican Republic continues to push for pretrial detentions in cases where no charges have been filed.

(Image credit: Ramon Espinosa)

The migrants hailed from 10 mostly Asian countries the U.S. has difficulty deporting migrants to directly. Panama is being used as a stopover.

(Image credit: Mark Schiefelbein)

If you're a back sleeper, a solid memory foam pillow -- or other similar options -- will help keep your spine and neck aligned.

Enjoy 10% off with Dell coupon, plus today's deals for laptops, monitors, and all things tech.

xAI's debut of the Grok 3 AI model allowed CEO Elon Musk and his team to show off the power of the latest iteration of Grok. Slightly lost in the benchmarking fireworks is what happens to the Grok 2 model, though. While Grok 3 is exclusive to Premium+ subscribers on X, Grok 2 will be even more accessible than before as it will shortly be open-source.

Grok 2 has been the default model for every Grok user, but the underlying engine will now have a chance to spread far further. Usually, AI models from big companies are locked down. You get access to them through paid subscriptions, and you can use them, but you can’t really see how they work or modify them to suit your specific needs. Open-sourcing Grok 2 means that developers will be able to dig into its code and customize it for use in entirely new applications.

Just how open the open-source version of Grok 2 will be is unclear, though. Sometimes, companies say open source even when they retain or hold back at least a little of their proprietary work. Still, Grok 2 is likely to have a second life in the hands of an open-source community that can tweak and experiment with it.

Despite its name, OpenAI has moved away from open-source. Its most powerful models remain proprietary. GPT-4, for example, remains locked down, with no open-source version in sight. Musk's deliberate contrast with OpenAI certainly appears, at least somewhat.

After stating his distrust of OpenAI CEO Sam Altman last fall, Musk highlighted his plan to open-source xAI models after completing the next iteration to contrast with ChatGPT's creator.

Worth noting that @xAI has been and will open source its models, including weights and everything, As we create the next version, we open source the prior version, as we did with Grok 1 when Grok 2 was released.October 4, 2024

Open thought

Not that xAI is unique in pursuing the open-source vision. Meta has dabbled with the open-source movement by releasing aspects of its Llama models since they first launched. The sudden superstar of AI, China-based DeepSeek, has also been pushing hard to open-source AI.

The open-sourcing of Grok 2 means they might lead to some intriguing AI tools popping up outside the usual walled gardens. We could see smaller companies and independent developers using it to create custom chatbots and personalized AI assistants, like an AI chatbot trained entirely on classic literature or one that helps indie video game developers generate lore for their worlds. They might prove you don’t need a multi-billion-dollar walled garden to develop something groundbreaking.

Then again, some of those creations will depend on what kind of limitations will be baked into the release. Even if xAI means it when it says Grok 2 will be open source, the company could still have its cake and eat it, too. The company can keep its cutting-edge products exclusive but let people have the older version to generate goodwill, get free innovation from independent developers, and maybe attract more people to sign up for Grok.

You might also like

• Grok steps out to mobile with new standalone iOS app
• Grok 3 launches today as Elon Musk's ‘scary good’ AI chatbot looks set to take on ChatGPT
• Grok-2 arrives on X with AI image creation, precious few guardrails, and lots of questions

Here are the answers for The New York Times Mini Crossword for Feb. 19.

It's the second tech company to agree to a payout after the Jan. 6 attack on the Capitol prompted Trump to be kicked off numerous social media platforms.

(Image credit: Jim Watson)

Grok 3 is "scary smart," according to Elon Musk, and is "an order of magnitude more powerful than Grok 2."

Former Brazilian President Jair Bolsonaro has been charged for allegedly helping to mastermind far right coup plot to upset the country's 2022 election result.

(Image credit: Eraldo Peres)

In some parts of the Plains region, the National Weather Service expects temperatures to reach as low as the negative 20s and 30s.

(Image credit: Joshua A. Bickel)

Jurors found A$AP Rocky, a rapper and the longtime partner of Rihanna, not guilty of firing a gun at a former friend on a Hollywood street in 2021. He could have faced up to 24 years in prison.

(Image credit: Damian Dovarganes)

President Trump campaigned on lowering the cost of IVF, after a controversial ruling made the treatment stop in Alabama for a short time

(Image credit: Sebastian Kaulitzki)

All consumer data will be deleted as of Feb. 28, so if you own one, download your data now.

NSF fired 168 employees, leaving the agency less equipped to fund a wide range of scientific research.

(Image credit: Narumon Bowonkitwanchai/Getty Images)

Malcolm in the Middle.'/>

For Frankie Muniz, balancing a career as a professional NASCAR driver and a TV actor isn't a fantasy; it'll be his very busy reality this upcoming year.

(Image credit: Mike Ehrmann)

Humane AI Pin's short, bumpy ride is over. HP announced on Tuesday (February 18) that it was acquiring the company and absorbing some employees for $116 million. According to a report in Bloomberg, the Human AI pin will be discontinued.

In a release on the acquisition, HP notes that it will "acquire key AI capabilities from Humane, including their AI-powered platform Cosmos, highly skilled technical talent, and intellectual property with more than 300 patents and patent applications." It makes no mention of hardware.

Humane AI was a fascinating wearable AI computing device that arrived in November 2023. Instead of a screen, the device featured a camera to watch your world and a tiny projector; you held out your hand and could see little messages and information from the pin. You interacted with the platform through voice and gestures. It sounded, at the time, like the future.

Humane AI Pin, though, ran aground in April 2024 when early adopters delivered scathing reviews. YouTube giant Marques Brownlee called it "The worst product I've ever reviewed... for now." That conclusion and a collection of other terrible reviews more or less sealed its fate. Well, that and issues with the batter clip that may have been a fire hazard. Even TechRadar rated it one of the biggest tech flops of 2024.

HP AI gets Humane

HP, though, sounds more than impressed with the system's AI capabilities and plans to weave them (or some form of them) throughout HP products. From the release: Humane's engineers "will help us create an intelligent ecosystem across all HP devices from AI PCs to smart printers and connected conference rooms. This will unlock new levels of functionality for our customers and deliver on the promises of AI."

For HP, which has been comparatively quiet on the AI innovation front, this acquisition could be a real boon to its business. As for the Humane AI team, they have nothing but praise for their new owners. In a release on the Humane website, Bethany Bongiorno and Imran Chaudhri, Co-founders of Humane, wrote, "We’re excited to join HP at such a pivotal moment in the industry and help shape the future of intelligent experiences,” said.

Humane's release also makes no mention of the AI Pin.

This is good news for Humane AI fans and, obviously, less good news for those who paid $699 for the pin. Perhaps HP will offer them, discounts on whatever HP products end up featuring Humane AI.

You might also like

• Humane AI Pin review roundup: an undercooked flop that's way ahead of its time
• Humane’s AI Pin is a screenless, wearable smartphone that’s straight out of Black Mirror

Hundreds of artists signed a letter sent to the National Endowment for the Arts asking it to reverse policy changes made as a result of recent executive orders issued by President Trump.

(Image credit: Graeme Sloan)

• iProov study finds older adults struggle most with deepfakes
• False confidence is widespread among the younger generation
• Social media is a deepfake hotspot, experts warn

As deepfake technology continues to advance, concerns over misinformation, fraud, and identity theft are growing, thanks to literacy in AI tools being at a startling low.

A recent iProov study claims most people struggle to distinguish deepfake content from reality, as it took 2,000 participants from the UK and US being exposed to a mix of real and AI-generated images and videos, finding only 0.1% of participants - two whole people - correctly distinguished between real and deepfake stimuli.

The study found older adults are particularly susceptible to AI-generated deception. Around 30% of those aged 55-64, and 39% of those over 65, had never heard of deepfakes before. While younger participants were more confident in their ability to detect deepfakes, their actual performance in the study did not improve.

Older generations are more vulnerable

Deepfake videos were significantly harder to detect than images, the study added,as participants were 36% less likely to correctly identify a fake video compared to an image, raising concerns about video-based fraud and misinformation.

Social media platforms were highlighted as major sources of deepfake content. Nearly half of the participants (49%) identified Meta platforms, including Facebook and Instagram, as the most common places where deepfakes are found, while 47% pointed to TikTok.

"[This underlines] how vulnerable both organizations and consumers are to the threat of identity fraud in the age of deepfakes," said Andrew Bud, founder and CEO of iProov.

"Criminals are exploiting consumers’ inability to distinguish real from fake imagery, putting personal information and financial security at risk."

Bud added even when people suspect a deepfake, most take no action. Only 20% of respondents said they would report a suspected deepfake if they encountered one online.

With deepfakes becoming increasingly sophisticated, iProov believes that human perception alone is no longer reliable for detection, and Bud emphasized the need for biometric security solutions with liveness detection to combat the threat of ever more convincing deepfake material.

“It’s down to technology companies to protect their customers by implementing robust security measures," he said. "Using facial biometrics with liveness provides a trustworthy authentication factor and prioritizes both security and individual control, ensuring that organizations and users can keep pace with these evolving threats."

You may also like

• Here's our rundown of the best endpoint protection around
• Take a look at the best AI video editors
• We've also listed the best social media management tools right now