Austria called on Monday for fresh efforts to regulate the use of AI in weapons systems that could create so-called 'killer robots', as it hosted a conference aimed at reviving largely stalled discussions on the issue. From a report: With AI technology advancing rapidly, weapons systems that could kill without human intervention are coming ever closer, posing ethical and legal challenges that most countries say need addressing soon. "We cannot let this moment pass without taking action. Now is the time to agree on international rules and norms to ensure human control," Austrian Foreign Minister Alexander Schallenberg told the meeting of non-governmental and international organisations as well as envoys from 143 countries.

"At least let us make sure that the most profound and far-reaching decision, who lives and who dies, remains in the hands of humans and not of machines," he said in an opening speech to the conference entitled "Humanity at the Crossroads: Autonomous Weapons Systems and the Challenge of Regulation." Years of discussions at the United Nations have produced few tangible results and many participants at the two-day conference in Vienna said the window for action was closing rapidly.

It's another speech-recognizing, AI-powered handheld device "about half the size of a phone," writes CNET. (Though the $199 device comes with a keyboard and a tiny 2.8-inch screen.) "The Rabbit R1 can identify items in its environment. Point it at a plant, and it can tell you what kind it is. Aim it at your lunch, and it can tell you what's in it.

"it also feels a bit like a novelty so far...." It can call an Uber, order dinner from Doordash, translate conversations, record voice memos, play songs from Spotify and more. Your phone can already do all of those things, but [CEO and founder Jesse] Lyu is promoting the Rabbit R1 as a faster and more natural way to do so... So far, the Rabbit R1 feels fun, fresh and interesting, but also frustrating at times. It intrigues me, but it also hasn't convinced me yet that there's room for another gadget in my life.... Many of the things it can do today feel smartphone-esque, like asking for the weather or playing songs on Spotify...

Visual search is the most interesting feature so far... It's pretty accurate for the most part so far. When I pointed it at my salad during lunch, it was able to tell me most of the ingredients. That's not what I asked. After all, who orders a dish without knowing what's in it? I asked the Rabbit R1 to tell me how many calories were in my lunch. While it couldn't provide the answer I wanted, I was impressed with its response... Overall, Rabbit R1's visual analysis worked pretty well for identifying things like plants and characters from pop culture. When describing my colleague's sneakers, the Rabbit R1 got the brand wrong...

So far, I've used the Rabbit R1 to take voice memos, translate speech from Spanish to English, and answer basic questions about things like weather forecasts. These features work as expected for the most part.
The article points out that Ray-Ban Meta Smart Glasses "also have multimodal AI, meaning the eyewear can 'see' what you see and tell you about it," and "you can already do something like this on your phone through Google's Gemini assistant on Android phones (or the Gemini section of the Google app for the iPhone).

"It's also very reminiscent of Google Lens, which has been around for years..."

An anonymous reader shared this report from the Associated Press: Tech giant Cisco Systems on Wednesday joined Microsoft and IBM in signing onto a Vatican-sponsored pledge to ensure artificial intelligence is developed and used ethically and to benefit the common good... The pledge outlines key pillars of ethical and responsible use of AI. It emphasizes that AI systems must be designed, used and regulated to serve and protect the dignity of all human beings, without discrimination, and their environments. It highlights principles of transparency, inclusion, responsibility, impartiality and security as necessary to guide all AI developments.

The document was unveiled and signed at a Vatican conference on Feb. 28, 2020... Pope Francis has called for an international treaty to ensure AI is developed and used ethically, devoting his annual peace message this year to the topic.

"A former high school athletic director was arrested Thursday morning," reports CBS News, "after allegedly using artificial intelligence to impersonate the school principal in a recording..." One-time Pikesville High School employee Dazhon Darien is facing charges that include theft, stalking, disruption of school operations and retaliation against a witness. Investigators determined he faked principal Eric Eiswert's voice and circulated the audio on social media in January. Darien's nickname, DJ, was among the names mentioned in the audio clips he allegedly faked, according to the Baltimore County State's Attorney's Office.

Baltimore County detectives say Darien created the recording as retaliation against Eiswert, who had launched an investigation into the potential mishandling of school funds, Baltimore County Police Chief Robert McCullough said on Thursday. Eiswert's voice, which police and AI experts believe was simulated, made disparaging comments toward Black students and the surrounding Jewish community. The audio was widely circulated on social media.
The article notes that after the faked recording circulated on social media the principal "was temporarily removed from the school, and waves of hate-filled messages circulated on social media, while the school received numerous phone calls."

The suspect had actually used the school's network multiple times to perform online searches for OpenAI tools, "which police linked to paid OpenAI accounts."

"Intel used to dominate the U.S. chip industry," writes CNBC. But now "it's struggling to stay relevant." Intel's long-awaited turnaround looks farther away than ever after the company reported dismal first-quarter earnings. Investors pushed the shares down 9% on Friday to their lowest level of the year. Although Intel's revenue is no longer shrinking and the company remains the biggest maker of processors that power PCs and laptops, sales in the first quarter trailed estimates. Intel also gave a soft forecast for the second quarter, suggesting weak demand... Intel is the worst-performing tech stock in the S&P 500 this year, down 37%.

Meanwhile, the two best-performing stocks in the index are chipmaker Nvidia and Super Micro Computer, which has been boosted by surging demand for Nvidia-based artificial intelligence servers. Intel, long the most valuable U.S. chipmaker, is now one-sixteenth the size of Nvidia by market cap. It's also smaller than Qualcomm, Broadcom, Texas Instruments, and AMD. For decades, it was the largest semiconductor company in the world by sales, but suffered seven straight quarters of revenue declines recently, and was passed by Nvidia last year.
Intel's problems "are decades in the making," according to CNBC, suggesting that one turning point was Apple's decision not to use Intel's chips in its iPhone. Now nearly every smartphone built uses Arm chips built by Apple and Qualcomm, while Apple's huge orders for TSMC chips "provided the cash to annually upgrade the manufacturing equipment at TSMC, which eventually surpassed Intel." Around 2017, mobile chips from Apple and Qualcomm started adding AI parts to their chips called neural processing units, another advancement over Intel's PC processors. The first Intel-based laptop with an NPU shipped late last year.

Intel has since lost share in its core PC chip business to chips that grew out of the mobile revolution... Apple stopped using Intel in its PCs in 2020. Macs now use Arm-based chips, and some of the first mainstream Windows laptops with Arm-based chips are coming out later this year. Low-cost laptops running Google ChromeOS are increasingly using Arm, too...

AMD made over 20% of server CPUs sold in 2022, and shipments grew 62% that year, according to an estimate from Counterpoint Research last year. AMD surpassed Intel's market cap the same year.

Sarah Perez reports via TechCrunch: EyeEm, the Berlin-based photo-sharing community that exited last year to Spanish company Freepik after going bankrupt, is now licensing its users' photos to train AI models. Earlier this month, the company informed users via email that it was adding a new clause to its Terms & Conditions that would grant it the rights to upload users' content to "train, develop, and improve software, algorithms, and machine-learning models." Users were given 30 days to opt out by removing all their content from EyeEm's platform. Otherwise, they were consenting to this use case for their work.

At the time of its 2023 acquisition, EyeEm's photo library included 160 million images and nearly 150,000 users. The company said it would merge its community with Freepik's over time. Despite its decline, almost 30,000 people are still downloading it each month, according to data from Appfigures. Once thought of as a possible challenger to Instagram -- or at least "Europe's Instagram" -- EyeEm had dwindled to a staff of three before selling to Freepik, TechCrunch's Ingrid Lunden previously reported. Joaquin Cuenca Abela, CEO of Freepik, hinted at the company's possible plans for EyeEm, saying it would explore how to bring more AI into the equation for creators on the platform. As it turns out, that meant selling their work to train AI models. [...]

Of note, the notice says that these deletions from EyeEm market and partner platforms could take up to 180 days. Yes, that's right: Requested deletions take up to 180 days but users only have 30 days to opt out. That means the only option is manually deleting photos one by one. Worse still, the company adds that: "You hereby acknowledge and agree that your authorization for EyeEm to market and license your Content according to sections 8 and 10 will remain valid until the Content is deleted from EyeEm and all partner platforms within the time frame indicated above. All license agreements entered into before complete deletion and the rights of use granted thereby remain unaffected by the request for deletion or the deletion." Section 8 is where licensing rights to train AI are detailed. In Section 10, EyeEm informs users they will forgo their right to any payouts for their work if they delete their account -- something users may think to do to avoid having their data fed to AI models. Gotcha!

An anonymous reader quotes a report from TechCrunch: Darktrace is set to go private in a deal that values the U.K.-based cybersecurity giant at around $5 billion. A newly formed entity called Luke Bidco Ltd., formed by private equity giant Thoma Bravo, has tabled an all-cash bid of $7.75 per share, which represents a 44% premium on its average price for the three-month period ending April 25. However, this premium drops to just 20% when juxtaposed against Darktrace's closing price Thursday, as the company's shares had risen 20% to 5.18 pounds in the past month.

Founded out of Cambridge, U.K., in 2013, Darktrace is best known for AI-enabled threat detection smarts, using machine learning to identify abnormal network activity and attempts at ransomware attacks, insider attacks, data breaches and more. The company claims big-name customers including Allianz, Airbus and the city of Las Vegas. After raising some $230 million in VC funding and hitting a private valuation of $1.65 billion, Darktrace went public on the London Stock Exchange in April 2021, with an opening-day valuation of $2.4 billion. Its shares hit an all-time high later that year of 9.45 pounds and plummeted to an all-time low of 2.29 pounds last February. But they had been steadily rising since the turn of the year and hadn't fallen below 4 pounds since the beginning of March.

The full valuation based on Thoma Bravo's offer amounts to $5.3 billion on what is known as a full-diluted basis, which takes into account all convertible securities and is designed to give a more comprehensive view of a company's valuation. However, the enterprise value in this instance is approximately $4.9 billion, which includes additional considerations such as debt and cash positions. [...] The deal is of course still subject to shareholder approval, but the companies said that they expect to complete the transaction by the end of 2024. "The proposed offer represents an attractive premium and an opportunity for shareholders to receive the certainty of a cash consideration at a fair value for their shares," Darktrace chair Gordon Hurst said. "The proposed acquisition will provide Darktrace access to a strong financial partner in Thoma Bravo, with deep software sector expertise, who can enhance the company's position as a best-in-class cyber AI business headquartered in the U.K."

Captchas that aim to distinguish humans from nefarious bots are demanding more brain power. WSJ: The companies and cybersecurity experts who design Captchas have been doing all they can to stay one step ahead of the bad actors figuring out how to crack them. A cottage industry of third-party Captcha-solving firms -- essentially, humans hired to solve the puzzles all day -- has emerged. More alarmingly, so has technology that can automatically solve the more rudimentary tests, such as identifying photos of motorcycles and reading distorted text. "Software has gotten really good at labeling photos," said Kevin Gosschalk, the founder and CEO of Arkose Labs, which designs what it calls "fraud and abuse prevention solutions," including Captchas. "So now enters a new era of Captcha -- logic based."

That shift explains why Captchas have started to both annoy and perplex. Users no longer have to simply identify things. They need to identify things and do something with that information -- move a puzzle piece, rotate an object, find the specter of a number hidden in a roomscape. Compounding this bewilderment is the addition to the mix of generative AI images, which creates new objects difficult for robots to identify but baffles humans who just want to log in. "Things are going to get even stranger, to be honest, because now you have to do something that's nonsensical," Gosschalk said. "Otherwise, large multimodal models will be able to understand."

An anonymous reader shares a report: With Windows 11 24H2 all geared up to have AI-intensive applications, Microsoft has added a code that will warn you if your PC does not meet the hardware requirements, according to code dug up by Twitter/X sleuth Albacore. The warning will be displayed as a watermark so you know that you cannot use certain AI-powered built-in apps because of an unsupported CPU.

404 Media: Apple has removed a number of AI image generation apps from the App Store after 404 Media found these apps advertised the ability to create nonconsensual nude images, a sign that app store operators are starting to take more action against these types of apps.

Overall, Apple removed three apps from the App Store, but only after we provided the company with links to the specific apps and their related ads, indicating the company was not able to find the apps that violated its policy itself.

Apple's action comes after we reported on Monday that Instagram advertises nonconsensual AI nude apps. By browsing Meta's Ad Library, which archives ads on its platform, when they ran, on what platforms, and who paid for them, we were able to find ads for five different apps, each with dozens of ads. Two of the ads were for web-based services, and three were for apps on the Apple App Store. Meta deleted the ads when we flagged them. Apple did not initially respond to a request for comment on that story, but reached out to me after it was published asking for more information. On Tuesday, Apple told us it removed the three apps on its App Store.

Sam Altman of OpenAI and the chief executives of Nvidia, Microsoft and Alphabet are among technology-industry leaders joining a new federal advisory board focused on the secure use of AI within U.S. critical infrastructure, in the Biden administration's latest effort to fill a regulatory vacuum over the rapidly proliferating technology. From a report: The Artificial Intelligence Safety and Security Board is part of a government push to protect the economy, public health and vital industries from being harmed by AI-powered threats, U.S. officials said. Working with the Department of Homeland Security, it will develop recommendations for power-grid operators, transportation-service providers and manufacturing plants, among others, on how to use AI while bulletproofing their systems against potential disruptions that could be caused by advances in the technology.

In addition to Nvidia's Jensen Huang, Microsoft's Satya Nadella, Alphabet's Sundar Pichai and other leaders in AI and technology, the panel of nearly two dozen consists of academics, civil-rights leaders and top executives at companies that work within a federally recognized critical-infrastructure sector, including Kathy Warden, chief executive of Northrop Grumman, and Delta Air Lines Chief Executive Ed Bastian. Other members are public officials, such as Maryland Gov. Wes Moore and Seattle Mayor Bruce Harrell, both Democrats.

An anonymous reader quotes a report from Tom's Hardware: TSMC announced its leading-edge 1.6nm-class process technology today, a new A16 manufacturing process that will be the company's first Angstrom-class production node and promises to outperform its predecessor, N2P, by a significant margin. The technology's most important innovation will be its backside power delivery network (BSPDN). Just like TSMC's 2nm-class nodes (N2, N2P, and N2X), the company's 1.6nm-class fabrication process will rely on gate-all-around (GAA) nanosheet transistors, but unlike the current and next-generation nodes, this one uses backside power delivery dubbed Super Power Rail. Transistor and BSPDN innovations enable tangible performance and efficiency improvements compared to TSMC's N2P: the new node promises an up to 10% higher clock rate at the same voltage and a 15%-20% lower power consumption at the same frequency and complexity. In addition, the new technology could enable 7%-10% higher transistor density, depending on the actual design.

The most important innovation of TSMC's A16 process, which was unveiled at the company's North American Technology Symposium 2024, is the introduction of the Super Power Rail (SPR), a sophisticated backside power delivery network (BSPDN). This technology is tailored specifically for AI and HPC processors that tend to have both complex signal wiring and dense power delivery networks. Backside power delivery will be implemented into many upcoming process technologies as it allows for an increase in transistor density and improved power delivery, which affects performance. Meanwhile, there are several ways to implement a BSPDN. TSMC's Super Power Rail plugs the backside power delivery network to each transistor's source and drain using a special contact that also reduces resistance to get the maximum performance and power efficiency possible. From a production perspective, this is one of the most complex BSPDN implementations and is more complex than Intel's Power Via. Volume production of A16 is slated for the second half of 2026. "Therefore, actual A16-made products will likely debut in 2027," notes the report. "This timeline positions A16 to potentially compete with Intel's 14A node, which will be Intel's most advanced node at the time."

Seagate has joined Western Digital in increasing the prices of hard drives, with rising demand due to the huge data requirements of AI taking the blame. AI is also behind a rapid growth in orders for Enterprise solid state drives. From a report: One of the big three makers of traditional rotating hard disk drives, Seagate informed customers that it is increasing prices effective immediately for new orders, but also for any changes to orders that are "over and above" previously committed volumes. This was disclosed in a letter from the company seen by analyst Trendforce, and comes just a couple of weeks after rival manufacturer Western Digital sent out a similar letter to customers informing them of price hikes.

According to Trendforce, the cause of the issue is two-fold: rising demand for high-capacity HDD products driven by the current craze for all things AI, and reduced production by hard drive manufacturers that means they are unable to meet the demand, leading to soaring prices. The rising demand comes from AI training requiring huge volumes of data: OpenAI's GPT-3 model is said to have been trained using 45TB of data, which may have been surpassed for newer models. And while flash-based SSDs boast high-speed and low-latency, storing everything in flash would still be costly. Seagate launched a 30TB hard drive line last year. Hard drive production was cut by as much as 20 percent over the last two years or so because of falling orders during the pandemic, and now manufacturers are unprepared for a sudden uptick in demand.

Thomas Claburn reports via The Register: Baltimore police have arrested Dazhon Leslie Darien, the former athletic director of Pikesville High School (PHS), for allegedly impersonating the school's principal using AI software to make it seem as if he made racist and antisemitic remarks. Darien, of Baltimore, Maryland, was subsequently charged with witness retaliation, stalking, theft, and disrupting school operations. He was detained late at night trying to board a flight at BWI Thurgood Marshall Airport. Security personnel stopped him because the declared firearm he had with him was improperly packed and an ensuing background check revealed an open warrant for his arrest.

"On January 17, 2024, the Baltimore County Police Department became aware of a voice recording being circulated on social media," said Robert McCullough, Chief of Baltimore County Police, at a streamed press conference today. "It was alleged the voice captured on the audio file belong to Mr Eric Eiswert, the Principal at the Pikesville High School. We now have conclusive evidence that the recording was not authentic. "The Baltimore County Police Department reached that determination after conducting an extensive investigation, which included bringing in a forensic analyst contracted with the FBI to review the recording. The results of the analysis indicated the recording contained traces of AI-generated content." McCullough said a second opinion from a forensic analyst at the University of California, Berkeley, also determined the recording was not authentic. "Based off of those findings and further investigation, it's been determined the recording was generated through the use of artificial intelligence technology," he said.

According to the warrant issued for Darien's arrest, the audio file was shared through social media on January 17 after being sent via email to school teachers. The recording sounded as if Principal Eric Eiswert had made remarks inflammatory enough to prompt a police visit to advise on protective security measures for staff. [...] The clip, according to the warrant, led to the temporary removal of Eiswert from his position and "a wave of hate-filled messages on social media and numerous calls to the school," and significantly disrupted school operations. Police say it led to threats against Eiswert and concerns about his safety. Eiswert told investigators that he believes the audio clip was fake as "he never had the conversations in the recording." And he said he believed Darien was responsible due to his technical familiarity with AI and had a possible motive: Eiswert said there "had been conversations with Darien about his contract not being renewed next semester due to frequent work performance challenges." "It is clear that we are also entering a new deeply concerning frontier as we continue to embrace emerging technology and its potential for innovation and social good," said John Olszewski, Baltimore County Executive, during a press conference. "We must also remain vigilant against those who would have used it for malicious intent. That will require us to be more aware and more discerning about the audio we hear and the images we see. We will need to be careful in our judgment."

The head of Indian IT company Tata Consultancy Services has said AI will result in "minimal" need for call centres in as soon as a year, with AI's rapid advances set to upend a vast industry across Asia and beyond. From a report: K Krithivasan, TCS chief executive, told the Financial Times that while "we have not seen any job reduction" so far, wider adoption of generative AI among multinational clients would overhaul the kind of customer help centres that have created mass employment in countries such as India and the Philippines. "In an ideal phase, if you ask me, there should be very minimal incoming call centres having incoming calls at all," he said. "We are in a situation where the technology should be able to predict a call coming and then proactively address the customer's pain point." He said chatbots would soon be able to analyse a customer's transaction history and do much of the work done by call centre agents. "That's where we are going...I don't think we are there today -- maybe a year or so down the line," he said.

Hartley Charlton reports via MacRumors: Apple is said to be developing its own AI server processor using TSMC's 3nm process, targeting mass production by the second half of 2025. According to a post by the Weibo user known as "Phone Chip Expert," Apple has ambitious plans to design its own artificial intelligence server processor. The user, who claims to have 25 years of experience in the integrated circuit industry, including work on Intel's Pentium processors, suggests this processor will be manufactured using TSMC's 3nm node.

Apple's purported move toward developing a specialist AI server processor is reflective of the company's ongoing strategy to vertically integrate its supply chain. By designing its own server chips, Apple can tailor hardware specifically to its software needs, potentially leading to more powerful and efficient technologies. Apple could use its own AI processors to enhance the performance of its data centers and future AI tools that rely on the cloud. While Apple is rumored to be prioritizing on-device processing for many of its upcoming AI tools, it is inevitable that some operations will have to occur in the cloud. By the time the custom processor could be integrated into operational servers in late 2025, Apple's new AI strategy should be well underway.

Axon on Tuesday announced a new tool called Draft One that uses artificial intelligence built on OpenAI's GPT-4 Turbo model to transcribe audio from body cameras and automatically turn it into a police report. Axon CEO Rick Smith told Forbes that police officers will then be able to review the document to ensure accuracy. From the report: Axon claims one early tester of the tool, Fort Collins Colorado Police Department, has seen an 82% decrease in time spent writing reports. "If an officer spends half their day reporting, and we can cut that in half, we have an opportunity to potentially free up 25% of an officer's time to be back out policing," Smith said. These reports, though, are often used as evidence in criminal trials, and critics are concerned that relying on AI could put people at risk by depending on language models that are known to "hallucinate," or make things up, as well as display racial bias, either blatantly or unconsciously.

"It's kind of a nightmare," said Dave Maass, surveillance technologies investigations director at the Electronic Frontier Foundation. "Police, who aren't specialists in AI, and aren't going to be specialists in recognizing the problems with AI, are going to use these systems to generate language that could affect millions of people in their involvement with the criminal justice system. What could go wrong?" Smith acknowledged there are dangers. "When people talk about bias in AI, it really is: Is this going to exacerbate racism by taking training data that's going to treat people differently?" he told Forbes. "That was the main risk."

Smith said Axon is recommending police don't use the AI to write reports for incidents as serious as a police shooting, where vital information could be missed. "An officer-involved shooting is likely a scenario where it would not be used, and I'd probably advise people against it, just because there's so much complexity, the stakes are so high." He said some early customers are only using Draft One for misdemeanors, though others are writing up "more significant incidents," including use-of-force cases. Axon, however, won't have control over how individual police departments use the tools.

Lenovo's latest ThinkPad P1 Gen 7 laptop is set to be the first to use the new LPCAMM2 memory form factor, the successor to SODIMM sticks. From a report: While Lenovo has largely focused on the AI performance of its new laptop, which is equipped with an Intel Core Ultra CPU and Nvidia RTX 3000 Ada GPU, the company also noted that its device was the first in the world to use the LPCAMM2 memory standard. LPCAMM2 uses 64 percent less space than SODIMM and 61 percent less active power, according to Lenovo. This is thanks to it being based on LPDDR5X memory instead of regular DDR5.

Designed specifically for laptops, the LPCAMM2 standard actually has its origins in tech developed by Dell. Simply termed CAMM (Compression Attached Memory Module), it first debuted as a proprietary type of memory in Dell's Precision 7670 in 2022. However, in 2023 the PC giant donated its intellectual property to JEDEC, the organization that standardizes memory technologies. CAMM became LPCAMM2 (Low-Power Compression Attached Memory Module) in September 2023 when JEDEC finally confirmed its specifications. Samsung promptly announced plans to produce LPCAMM2 sticks, and claimed they would have 50 percent more performance and 70 percent more efficiency than their SODIMM-based predecessors. Plus, LPCAMM2 can offer dual-channel memory without requiring a second module.

Adobe researchers have developed a new generative AI model called VideoGigaGAN that can upscale blurry videos at up to eight times their original resolution. From a report: Introduced in a paper published on April 18th, Adobe claims VideoGigaGAN is superior to other Video Super Resolution (VSR) methods as it can provide more fine-grained details without introducing any "AI weirdness" to the footage. In a nutshell, Generative Adversarial Networks (GANs) are effective for upscaling still images to a higher resolution, but struggle to do the same for video without introducing flickering and other unwanted artifacts. Other upscaling methods can avoid this, but the results aren't as sharp or detailed. VideoGigaGAN aims to provide the best of both worlds -- the higher image/video quality of GAN models, with fewer flickering or distortion issues across output frames. The company has provided several examples here that show its work in full resolution.

Just as Google, Samsung and Microsoft continue to push their efforts with generative AI on PCs and mobile devices, Apple is moving to join the party with OpenELM, a new family of open source large language models (LLMs) that can run entirely on a single device rather than having to connect to cloud servers. From a report: Released a few hours ago on AI code community Hugging Face, OpenELM consists of small models designed to perform efficiently at text generation tasks. There are eight OpenELM models in total -- four pre-trained and four instruction-tuned -- covering different parameter sizes between 270 million and 3 billion parameters (referring to the connections between artificial neurons in an LLM, and more parameters typically denote greater performance and more capabilities, though not always).

[...] Apple is offering the weights of its OpenELM models under what it deems a "sample code license," along with different checkpoints from training, stats on how the models perform as well as instructions for pre-training, evaluation, instruction tuning and parameter-efficient fine tuning. The sample code license does not prohibit commercial usage or modification, only mandating that "if you redistribute the Apple Software in its entirety and without modifications, you must retain this notice and the following text and disclaimers in all such redistributions of the Apple Software." The company further notes that the models "are made available without any safety guarantees. Consequently, there exists the possibility of these models producing outputs that are inaccurate, harmful, biased, or objectionable in response to user prompts."