Tag: deepmind

Google DeepMind to Use SynthID to Watermark Gemini and Veo’s AI-Generated Content

Google made a large number of artificial intelligence (AI)-based announcements late Tuesday during its I/O 2024 keynote session. These include new AI models, upgrades to existing foundation models, integration of AI features into Google’s products, and more. The tech giant also focused on AI safety and expanded the usage of its native watermarking technology for AI-generated content, dubbed SynthID. This new toolkit will now be embedding watermarks for text generated by the Gemini app and web client, and videos generated by Veo.

SynthID was first unveiled by Google DeepMind in August 2023 as a beta project aimed at correctly labelling AI-generated content. The need for such a solution was felt due to the rise of instances where these synthetically created media were shared as real. These were used to spread misinformation and cybercrimes such as phishing. The tech giant first used this technology in November 2023, when it was used to watermark AI-generated audio created through its Lyria model. The toolkit added watermarks as a waveform to the audio to make it imperceptible yet detectable.

Now, Google is expanding the usage of SynthID to include text and video generation. It will now watermark the text generated using the Gemini app and the website. For this, the toolkit will target the generation process itself. Every text-based AI model uses tokens — which can be words, syllables, or phrases — to train. The training process also includes understanding the flow of using these tokens, or the order the tokens should follow to generate the most coherent response.

SynthID introduces “additional information in the token distribution at the point of generation by modulating the likelihood of tokens being generated.” This way it assigns a number to certain words in a block of generated text. When detecting whether AI was used to generate the text, it checks the score against its adjusted probability scores to determine whether the source could be an AI model. DeepMind highlighted in a post that this technique is useful when an AI generates long creative text as it is easier for probability models to check how it was created. However, for shorter factual responses, the detection may not be as accurate.

The company is also expanding SynthID to recently unveiled Veo’s AI-generated videos. Google said the technology will embed watermarks directly into the pixels of every video frame which will be imperceptible to the human eye but will show up when a detection system is used.

In the coming months, Google plans to open-source SynthID text watermarking through its Responsible Generative AI toolkit. It will also publish a detailed research paper explaining the text watermarking technology.

Check out our Latest News and Follow us at Facebook

Original Source

Google DeepMind Scientists Said to Be in Talks to Leave and Form AI Startup

A pair of scientists at Google DeepMind, the Alphabet artificial intelligence division, have been talking with investors about forming an AI startup in Paris, according to people familiar with the conversations. The team has held discussions with potential investors about a financing round that may exceed EUR 200 million (roughly Rs. 1,800 crore) — a large sum, even for the buzzy field of AI, the people said. Laurent Sifre, who has been working as a scientist at DeepMind, is in talks to form the company, known at the moment as Holistic, with fellow DeepMind scientist Karl Tuyls, said the people, asking not to be identified discussing private information. They said the venture may be focused on building a new AI model.

Sifre and Tuyls didn’t respond to multiple requests for comment. A DeepMind representative declined to comment on the startup plans. The two have given notice to leave the company, a person familiar with the situation said, asking not to be identified because the information is private.

Sifre was a co-author of the 2016 DeepMind research on Go, a seminal work that showed a computer system beating masters of the ancient game for the first time, which sparked an international frenzy over AI. Tuyls has worked on research into game theory and multi-agent reinforcement learning, a branch of AI that explores interactions between autonomous actors, often through video games.

Both Sifre and Tuyls are widely considered leaders in their field, and the unusually large financing round being discussed is further evidence of strong investor interest in the technology. That’s particularly true in France, where venture capitalists and business tycoons have poured funds into startups emerging from Parisian universities and the AI hubs of Silicon Valley firms.

Mistral AI, an OpenAI rival whose chief executive officer also worked for DeepMind, formed in early 2023 and had raised two sizable rounds by the year’s end to earn a valuation around $2 billion (roughly Rs. 16,600). Kyutai, a nonprofit AI research lab, was formed in November with EUR 300 (roughly Rs. 2,700 crore) million in initial funding.

The new French startup under discussion is different from Holistic AI, an enterprise software business based in London.

Affiliate links may be automatically generated – see our ethics statement for details.

Check out our Latest News and Follow us at Facebook

Original Source

Alphabet to Consolidate Google Brain, DeepMind AI Research Units in Race to Keep Up With Rival ChatGPT

Alphabet is combining Google Brain and DeepMind, as it doubles down on artificial intelligence research in its race to compete with rival systems like OpenAI’s ChatGPT chatbot.

The new division will be led by DeepMind CEO Demis Hassabis and its setting up will ensure “bold and responsible development of general AI”, Alphabet CEO Sundar Pichai said in a blog post on Thursday.

Alphabet said the teams that are being combined have delivered a number of high-profile projects including the transformer, technology that formed the bedrock of some of OpenAI’s own work.

Going forward, the Alphabet staff will work on “multimodal” AI, like OpenAI’s latest model GPT-4, which can respond not only to text prompts but to image inputs as well to generate new content. Google has for decades dominated the search market, with a share of over 80 percent, but Wall Street fears that the Alphabet unit could fall behind Microsoft Corp in the fast-moving AI race. Technology from OpenAI, funded by Microsoft, powers the rival software maker’s updated Bing search engine.

Alphabet announced the launch of Bard in February to take on ChatGPT as well. It lost $100 billion in value on Feb. 8 after Bard shared inaccurate information in a promotional video and a company event failed to dazzle.

Alphabet shares were up 2 percent on Thursday. Earlier this week, it was reported that Alphabet shares fell over 4 percent in premarket trading after a report that said South Korea’s Samsung Electronics was considering replacing Google with Microsoft-owned Bing as the default search engine on its devices.

The report, published by the New York Times over the weekend, underscored the growing challenges Google’s $162-billion (roughly Rs. 13,29,477 crore) a-year search engine business faces from Bing — a minor player that has risen in prominence recently after the integration of the artificial intelligence tech behind ChatGPT.

Google’s reaction to the threat was “panic” as the company earns an estimated $3 billion (roughly Rs. 24,625 crore) in annual revenue from the Samsung contract, the report said, citing internal messages.

© Thomson Reuters 2023

Affiliate links may be automatically generated – see our ethics statement for details.

Check out our Latest News and Follow us at Facebook

Original Source

Alphabet’s DeepMind Predicts Nearly All Protein Structures Known to Science

Alphabet’s DeepMind unveiled its AlphaFold software in 2020, and it has since been used to create the AlphaFold Protein Structure Database (AlphaFold DB). It includes highly accurate protein structures predicted by the software. This database has been accessed by researchers to tackle real-world problems like plastic pollution, antibiotic resistance, and more. DeepMind has now partnered up with European Molecular Biology Laboratory (EMBL) to release predicted structures of nearly all catalogued proteins known to science. The AlphaFold DB now reportedly holds over 200 million structures. The applications of these findings across various fields of science are endless.

DeepMind revealed on Thursday that AlphaFold has predicted nearly all proteins known to science. The new predicted proteins include structures of plants, bacteria, animals, and other organisms. These could open new avenues for researchers that are attempting to tackle important issues like sustainability, food insecurity, and neglected diseases.

“…AlphaFold has already accelerated and enabled massive discoveries, including cracking the structure of the nuclear pore complex. And with this new addition of structures illuminating nearly the entire protein universe, we can expect more biological mysteries to be solved each day, ” said Eric Topol, Founder and Director of the Scripps Research Translational Institute.

The AlphaFold DB works similar to the Google search engine for researchers. It provides them instant access to predicted models of proteins. These models have been reportedly cited in important research like finding a cure for Parkinson’s Disease and developing a malaria vaccine.

CEO of DeepMind Demis Hassabis believes, “…[that] AI might turn out to be just the right technique to cope with the dynamic complexity of biology.” According to the company, over 500,000 researchers across 190 countries have accessed AlphaFold DB. The company has also partnered up with the Drugs for Neglected Diseases Initiative (DNDI) for “finding life-saving cures for diseases like Leishmaniasis and Chagas disease that disproportionately affect people in poorer parts of the world.”

Check out our Latest News and Follow us at Facebook

Original Source

Exit mobile version