Month: May 2024
Mistral AI introduces its first LLM for coding, fluent in 80 programming languages
Mistral AI has confirmed the launch of Codestral, its first LLM designed to assist developers write code. The Amazon-backed French… Continue reading Mistral AI introduces its first LLM for coding, fluent in 80 programming languages
The post Mistral AI introduces its first LLM for coding, fluent in 80 programming languages appeared first on ReadWrite.
Mistral AI has confirmed the launch of Codestral, its first LLM designed to assist developers write code.
The Amazon-backed French company is known for its Mistral and Mixtral language models, but Codestral appears set to raise the bar higher. It is trained in over 80 programming languages including Python, Java, C, C++, JavaScript, Bash, Swift, and Fortran, which can be traced back to 1957.
Operating in English, the LLM will be able to complete coding tasks, write tasks, and use a ‘fill-in’ mechanism for partial code. The program has also been tipped to reduce the risk of errors and bugs in code, for those companies and developers who utilize Codestral.
With a valuation of £6 billion, Mistral AI is making waves in the industry just over a year since the startup was formed as a “European rival to OpenAI”
In an official blog post, the generative AI model for coding was described as “open weight” but that could be contested. The license prohibits Codestral’s use for any commercial purposes but there is a clause for “development” with a further catch as the terms rule out “any internal usage by employees in the context of the company’s business activities.”
Today we are releasing Codestral-22B, our first code model! Codestral is trained on more than 80 programming languages and outperforms the performance of previous code models, including the largest ones.
It is available on our API platform, through instruct and… pic.twitter.com/YS85u0oIfG
— Guillaume Lample @ ICLR 2024 (@GuillaumeLample) May 29, 2024
Codestral’s encouraging benchmark performance
Mistral compared Codestral 22B to CodeLlama 70B, DeepSeek Coder 33B, and Llama 3 70B. Fewer parameters are deployed by Codestral but it appeared to perform favorably on Python, SQL, and other programming languages according to various benchmarks conducted. It also has a greater 32k context window.
Mistral AI has made an impression despite its fledgling status in the AI sector. As well as gaining backing from Amazon, Microsoft agreed terms with the company in February to use its models on the Azure platform. The deal also enables Azure AI Studio customers to avail of Mistral Large, the premium offering, and the most powerful, from the French startup.
If you want a have a closer look at Codestral, navigate to Le Chat, and under the prompt box, press the ‘model’ drop down and select Codestral.
Image credit: Ideogram
The post Mistral AI introduces its first LLM for coding, fluent in 80 programming languages appeared first on ReadWrite.
New York Governor To Launch Bill Banning Smartphones in Schools
The New York governor, Kathy Hochul, plans to introduce a bill banning smartphones in schools, the latest in a series of legislative moves aimed at online child safety by New York’s top official. From a report: “I have seen these addictive algorithms pull in young people, literally capture them and make them prisoners in a space where they are cut off from human connection, social interaction and normal classroom activity,” she said. Hochul said she would launch the bill later this year and take it up in New York’s next legislative session, which begins in January 2025. If passed, schoolchildren will be allowed to carry simple phones that cannot access the internet but do have the capability to send texts, which has been a sticking point for parents. She did not offer specifics on enforcing the prohibition. “Parents are very anxious about mass shootings in school,” she said. “Parents want the ability to have some form of connection in an emergency situation.” The smartphone-ban bill will follow two others Hochul is pushing that outline measures to safeguard children’s privacy online and limit their access to certain features of social networks.
Read more of this story at Slashdot.
The New York governor, Kathy Hochul, plans to introduce a bill banning smartphones in schools, the latest in a series of legislative moves aimed at online child safety by New York’s top official. From a report: “I have seen these addictive algorithms pull in young people, literally capture them and make them prisoners in a space where they are cut off from human connection, social interaction and normal classroom activity,” she said. Hochul said she would launch the bill later this year and take it up in New York’s next legislative session, which begins in January 2025. If passed, schoolchildren will be allowed to carry simple phones that cannot access the internet but do have the capability to send texts, which has been a sticking point for parents. She did not offer specifics on enforcing the prohibition. “Parents are very anxious about mass shootings in school,” she said. “Parents want the ability to have some form of connection in an emergency situation.” The smartphone-ban bill will follow two others Hochul is pushing that outline measures to safeguard children’s privacy online and limit their access to certain features of social networks.
Read more of this story at Slashdot.
OpenAI has a has a new version of ChatGPT just for universities
OpenAI is bringing ChatGPT to college campuses across the country. On Thursday, the company announced ChatGPT Edu, a version of ChatGPT built specifically for students, academics, faculty. “ChatGPT Edu is designed for schools that want to deploy AI more broadly to students and their campus communities,” the company said in a blog post.
ChatGPT Edu includes access to GPT-4o, OpenAI’s latest large language model that the company revealed earlier this month. OpenAI claims that the model is much better than its previous versions at interpreting text, coding, and mathematics, analyzing data sets, and being able to access the web. ChatGPT Edu will also have significantly higher message limits than the free version of ChatGPT and allow universities to build custom versions of ChatGPT trained on their own data — confusingly called GPTs — and share them within university workspaces. OpenAI claims that conversations and data from ChatGPT Edu won’t be used to train OpenAI’s models.
Although the introduction of ChatGPT in late 2022 initially raised concerns about academic integrity and potential misuse in educational environments, universities have increasingly been experimenting with using generative AI for both teaching as well as research. OpenAI said that it built ChatGPT Edu after it saw Wharton, Arizona State University and Columbia among others using ChatGPT Enterprise.
MBA undergrads at Wharton, for instance, completed their final reflection assignments by training a GPT trained on course materials have having discussions with the chatbot, while Arizona State University is experimenting with its own GPTs that engage German conversations with students learning the language.This article originally appeared on Engadget at https://www.engadget.com/openai-has-a-has-a-new-version-of-chatgpt-just-for-universities-191350708.html?src=rss
OpenAI is bringing ChatGPT to college campuses across the country. On Thursday, the company announced ChatGPT Edu, a version of ChatGPT built specifically for students, academics, faculty. “ChatGPT Edu is designed for schools that want to deploy AI more broadly to students and their campus communities,” the company said in a blog post.
ChatGPT Edu includes access to GPT-4o, OpenAI’s latest large language model that the company revealed earlier this month. OpenAI claims that the model is much better than its previous versions at interpreting text, coding, and mathematics, analyzing data sets, and being able to access the web. ChatGPT Edu will also have significantly higher message limits than the free version of ChatGPT and allow universities to build custom versions of ChatGPT trained on their own data — confusingly called GPTs — and share them within university workspaces. OpenAI claims that conversations and data from ChatGPT Edu won’t be used to train OpenAI’s models.
Although the introduction of ChatGPT in late 2022 initially raised concerns about academic integrity and potential misuse in educational environments, universities have increasingly been experimenting with using generative AI for both teaching as well as research. OpenAI said that it built ChatGPT Edu after it saw Wharton, Arizona State University and Columbia among others using ChatGPT Enterprise.
MBA undergrads at Wharton, for instance, completed their final reflection assignments by training a GPT trained on course materials have having discussions with the chatbot, while Arizona State University is experimenting with its own GPTs that engage German conversations with students learning the language.
This article originally appeared on Engadget at https://www.engadget.com/openai-has-a-has-a-new-version-of-chatgpt-just-for-universities-191350708.html?src=rss
The TweetDeck clone for Threads is finally here
The web version of Threads just got a whole lot more usable. Meta is rolling out the new Tweetdeck-like column view to all users, after it started testing the feature earlier this month.
The new look, which some Threads users have nicknamed “ThreadsDeck,” allows you to pin up to 100 different feeds to the Threads home page. Each column can also be set to auto-update so you can follow new posts as they come in.
While Tweetdeck fans will appreciate the familiarity, the new view also addresses some of the headaches of Threads’ “for you” algorithm. While the new column layout won’t let you eliminate the “for you” feed entirely, you can more easily hide it in the column view and put your “following” feed and any others side-by-side. It also vastly increases the number of posts you can view at any one time, making it a much more practical view for finding real-time information.
Threads’ “for you” feed has been a long-running source of complaints from many users, who question why the feed often seems to surface days-old posts and random stories from unconnected strangers instead of more timely updates. Meta’s controversial decision to automatically limit “political” content in its recommendations is also often criticized.
At the same time, the company does seem to recognize that users want to be able to find news and other timely updates. Threads has also recently gotten a “recent” filter for search results, and has a limited version of “trending topics.” Threads’ column view also makes both of these features more visible since you can set dedicated columns for specific search topics to track at-a-glance.This article originally appeared on Engadget at https://www.engadget.com/the-tweetdeck-clone-for-threads-is-finally-here-191132892.html?src=rss
The web version of Threads just got a whole lot more usable. Meta is rolling out the new Tweetdeck-like column view to all users, after it started testing the feature earlier this month.
The new look, which some Threads users have nicknamed “ThreadsDeck,” allows you to pin up to 100 different feeds to the Threads home page. Each column can also be set to auto-update so you can follow new posts as they come in.
While Tweetdeck fans will appreciate the familiarity, the new view also addresses some of the headaches of Threads’ “for you” algorithm. While the new column layout won’t let you eliminate the “for you” feed entirely, you can more easily hide it in the column view and put your “following” feed and any others side-by-side. It also vastly increases the number of posts you can view at any one time, making it a much more practical view for finding real-time information.
Threads’ “for you” feed has been a long-running source of complaints from many users, who question why the feed often seems to surface days-old posts and random stories from unconnected strangers instead of more timely updates. Meta’s controversial decision to automatically limit “political” content in its recommendations is also often criticized.
At the same time, the company does seem to recognize that users want to be able to find news and other timely updates. Threads has also recently gotten a “recent” filter for search results, and has a limited version of “trending topics.” Threads’ column view also makes both of these features more visible since you can set dedicated columns for specific search topics to track at-a-glance.
This article originally appeared on Engadget at https://www.engadget.com/the-tweetdeck-clone-for-threads-is-finally-here-191132892.html?src=rss
Tesla aims to launch its FSD software in China
Image: Owen Grove / The Verge
Tesla is trying to get its advanced driver assistant “Full Self-Driving” (FSD) software registered with China’s Ministry of Industry and Information Technology, according to sources speaking to Reuters. It’s a preliminary step that could lead to employee testing of the system before it eventually rolls out to customers in Tesla’s second-largest global market.
The news comes after CEO Elon Musk’s visit to China last month when government officials lifted restrictions on Tesla vehicle use, and the company secured a deal to collect mapping data. FSD currently costs $99 per month in the US, or $8,000, for a one-time activation, and according to Reuters’ sources, in China, it could have a similar monthly subscription fee. It will also face competition from Chinese automakers like BYD, Nio, and Xpeng, which already have advanced driver-assistance systems available in the country.
The Tesla FSD system currently available in China is a less capable iteration of the software, similar to the company’s Autopilot system that can assist with steering but doesn’t automatically navigate city streets. In the US, a report released by NHTSA in April linked FSD and Autopilot to hundreds of crashes and dozens of deaths.
Tesla has also not yet launched FSD in Europe. As we wrote in 2022, “To obtain that approval, Tesla must convincingly demonstrate that cars with FSD are at least as safe as those without it. At least so far, it hasn’t.”
Image: Owen Grove / The Verge
Tesla is trying to get its advanced driver assistant “Full Self-Driving” (FSD) software registered with China’s Ministry of Industry and Information Technology, according to sources speaking to Reuters. It’s a preliminary step that could lead to employee testing of the system before it eventually rolls out to customers in Tesla’s second-largest global market.
The news comes after CEO Elon Musk’s visit to China last month when government officials lifted restrictions on Tesla vehicle use, and the company secured a deal to collect mapping data. FSD currently costs $99 per month in the US, or $8,000, for a one-time activation, and according to Reuters’ sources, in China, it could have a similar monthly subscription fee. It will also face competition from Chinese automakers like BYD, Nio, and Xpeng, which already have advanced driver-assistance systems available in the country.
The Tesla FSD system currently available in China is a less capable iteration of the software, similar to the company’s Autopilot system that can assist with steering but doesn’t automatically navigate city streets. In the US, a report released by NHTSA in April linked FSD and Autopilot to hundreds of crashes and dozens of deaths.
Tesla has also not yet launched FSD in Europe. As we wrote in 2022, “To obtain that approval, Tesla must convincingly demonstrate that cars with FSD are at least as safe as those without it. At least so far, it hasn’t.”