close
close

Meta's lama was downloaded over a billion times

Welcome back to the request.

Meta announced today lamaHis open source model with great language has seen over A billion downloads Since its publication in 2023. The company used the milestone to highlight some of the business applications of its model, including personalization Recommendations for Spotify and easier M&A transactions. Mark Zuckerberg, CEO of Meta, celebrated the performance by publishing a gif of a jumping lama.

Now let's get in the headlines.


Big games

Google Deepmind Called the introduction of two new AI models for robots known last week. The first is Gemini RoboticsA model “Vision language action” based on Gemini 2.0. The second is Gemini Robotics“A Gemini model with an advanced spatial understanding that enables robotics to carry out their own programs,” said the company. Deepmind said that it is a partnership with Apptronik with Humanoid Robotics Company to use the model in a new series of robots.


Chip wars

Intel's new CEO Lip-Bu Tan Reuters plans to make major changes to the way the chip manufacturer does business. This includes Personnel cuts on middle management In order to accelerate the processes and aggressive efforts to advertise new customers to his foundry that creates customer -specific chips for Amazon and Microsoft. According to reports, Tan plans to design Intel and Produce new chips Supply AI servers with electricity.


Future of work

When people take more AI tools in their work unpredictable ways. A typical example: Wired reports that a developer who used Cursor ai To produce code, the AI ​​assistant was disabled blamed him And refused to generate more. It told the developer that he should Code the project itself so that he could better maintain the program. This is not the first time that an AI assistant refused to perform a task: last year, Openai had to publish an update to Chatgpt-4 to fix its Problem with “laziness” Either to return very simple results or to refuse to answer commands. Maybe we have to say our AI assistant “please” more often in the future?


Data

Openai plans a beta test for a new function for its Chatgpt team Subscribers who would connect the LLM To your Google Drive and Slack So that his chatbot can answer questions that are informed by internal documents and discussions Techcrunch. According to reports, the company is planning to expand this function to expand further systems in the future, such as: Box and Microsoft SharePoint. The new connection function is powered by a custom GPT-4O model.


Ki deal of the week

Insilico medicineand uses AI to develop new medication and collected a round of Series E of 110 million US Over 1 billion US dollars. The company said that it would use the capital for the further development of its 30 drug candidates discovered by AI and its models. Insilico currently has a AI-discovered medication for lung disease in lung fibrosis in human studies.


Deep dive

Rabbi Yitzi Hurwitz has spent a decade to only communicate with his eyes. Diagnosed Amyotrophic lateral sclerosis (As), also known as “Lou Gehrig disease” in 2013, the quick loss of muscle control meant that he can only “speak” Say words With an eye diagram. It is as frustrating and demoralizing as you can imagine.

One of the 30,000 Americans Hurwitz currently lives with (about 5,000 new cases are diagnosed every year). Among them one is developed by Andreas Forsland, CEO von, knowledge. It is a Brain-computer interface (BCI) who can help paralyzed patients interact and communicate with computers. And in contrast to similar technologies of Elon Musks NeuralinkIt does not require surgical implantation in the skull. The company announced last week that it started its first clinical study in which the technology is examined with 10 as patients. Rabbi Hurwitz is one and has been training on the device three days a week.

Hurwitz 'caregiver said forbes that they are already See progress. “It looks very promising,” they said. “When he opened the keyboard for the first time, he actually managed to say something himself, and that was surprising.

Knowledge has raised $ 25 million Of venture companies such as Prime Movers Lab and Amazon Alexa Fund to develop its BCI device, which is known as Axon-R. It is a helmet that can pursue brain waves with an EEG and eye movements, which means that the users can interact with an augmented reality display. This enables a variety of interactions, including the use of the device for “entering” words that are spoken loudly via a computer speaker. The company used Generative AI models This train about the patients' Own language patternAnd over time, it adapts to you, which should make communication faster.

Read more at Forbes.


Model behavior

One of the first things that children do when they come Kindergarten is to learn how to do it Tell the time. This slightly developed ability is something that many Multimodal KI models According to a new study by researchers from the University of Edinburgh, a new study is still fighting. Their results show that even state -of -the -art models failed to do the clock positions correctly more than 25% of cases and did worse with watches that were used more stylized or applied Roman digits.

More at Forbes

ForbesTesla is no longer Elon Musk's most valuable capitalForbesHow Pam Bondi became rich in just six years outside the officeForbesMeet the influencer to help small business owners stay healthy