| Here's a look at today's daily AI briefing: - Tech companies sign pledge to tackle AI election risks.
- New chip uses light instead of electricity for AI computing.
- Softbank founder seeks $100B for AI chip venture.
- Journal retracts study featuring incorrect AI-generated images.
Beth
p/beth-duckett | |
| 1 | | Twenty tech giants, including OpenAI, Amazon, and Google, have joined forces to work to prevent AI from deceiving voters and disrupting global elections this year. Unveiled at today's Munich Security Conference, the voluntary "Tech Accord to Combat Deceptive Use of AI in 2024 Elections" will help fight deepfakes that can mimic politicians and others before elections. More: - Other signatories include Adobe, TikTok, IBM, Anthropic, Inflection AI, Meta, Snap, and X, formerly Twitter.
- The companies agreed to work together on tools to detect deceptive AI-generated media, voter education campaigns, and technologies such as watermarking and metadata embedding.
- Overall, they agreed to eight key commitments, including checking for risks in their models, dealing with fake content on their platforms, and being transparent about it.
- The announcement clarifies that the commitments only apply to the services each company offers if they're relevant.
Zoom out: - At least 83 elections are scheduled around the world this year, the highest concentration expected in the next 24 years.
- According to an Axios-Generation Lab-Syracuse University survey, top computer science professors believe that AI's audio and video deepfake capabilities are now at an advanced level.
- When asked about how to fight deepfake misuse, the professors chose digital authenticity certification as the most effective approach (39%), followed by public education on deepfakes (29%), improved detection technology (18%), and stricter regulations (13%).
|     | |
| 2 | | University of Pennsylvania engineers created a silicon-photonic chip that uses light instead of electricity for AI computing. The chip could improve AI model training by speeding up data transfer and cutting down on electricity consumption. The design is market-ready and adaptable for GPUs as an add-on. More: - The researchers created a platform for more efficient vector-matrix multiplication, a key math operation for neural network training.
- The chip has thinner silicon in specific areas, creating variations in height that can control where light goes through the chip.
- This scatters the light in precise patterns, allowing the chip to execute math calculations at light speed.
- The researchers worked with a commercial foundry to produce the SiPh chips, adjusting the design to match market-available sizes to make them ready for deployment.
- The study was published today in Nature Photonics journal.
| | | |
| 3 | | SoftBank founder Masayoshi Son is looking to secure up to $100B to build an AI chip venture to rival Nvidia, sources told Bloomberg. The venture, code-named Izanagi, would produce semiconductors that power AI technology. More: - According to the sources, Son is aiming for the company to create an "AI chip powerhouse" that complements its chip designer, Arm.
- SoftBank Group owns roughly 90% of Arm.
- Under one plan, SoftBank could contribute $30B to the venture, while institutions from the Middle East could supply the other $70B.
- Son named the project "Izanagi," after the Japanese god of life and creation, to reflect its focus on artificial general intelligence (AGI).
- The specifics of the project, including funding, are still up in the air.
| | | |
| 4 | | A scientific journal has retracted a study containing misleading and anatomically incorrect images created with the AI art generator Midjourney. One of the three AI-created images featured gibberish text along with the image of a rat with unusually large genitals. More: - The paper, written by three researchers in China, summarized recent research on sperm stem cells.
- A researcher from India edited the paper, which was reviewed by two individuals based in the U.S. and India.
- The peer-reviewed journal, "Frontiers in Cell Development and Biology," has retracted the study entirely, saying it did not meet its standards for "editorial and scientific rigor."
- The journal also apologized on its website, clarifying that although one reviewer raised concerns about the images, they were ignored.
- Frontiers' policies allow the use of generative AI as long as it's disclosed and accurate.
| | | |
| 5 | | Virgin Group's Richard Branson and Charles Oppenheimer, J. Robert Oppenheimer's grandson, signed an open letter calling on world leaders to take action against uncontrolled AI and other world threats. The letter, published Thursday and shared with world governments, accuses leaders of not responding to the threats with "the wisdom and urgency required." More: - Other signatories include ex-U.N. Secretary-General Ban Ki-moon, singer Annie Lennox, former Unilever CEO Paul Polman, former Irish president Mary Robinson, and former Colombian president Juan Manuel Santos.
- The letter comes from The Elders, an NGO founded by Nelson Mandela to promote global human rights and peace.
- It's also being backed by the Future of Life Institute, the nonprofit focused on risks from advanced technologies, which last year called for a pause on all AI development that's more advanced than GPT-4.
Zoom out: - Over 130 people have now signed the letter urging action against unregulated AI along with climate change, pandemics, and nuclear arms.
- The message asks leaders for a long-view strategy and "determination to resolve intractable problems, not just manage them, the wisdom to make decisions based on scientific evidence and reason, and the humility to listen to all those affected."
| | | |
| 6 | | Weekly roundup — The top AI stories you might have missed this week: - OpenAI announced Sora, its new AI system that generates high-definition videos from text prompts. Sora, the Japanese word for sky, can create realistic-looking videos of up to a minute long. It will first launch to "a limited number of creators."
- ChatGPT is undergoing testing for a long-term "memory" feature. Users can tell ChatGPT to remember specific details, or the bot can gradually pick up and store data about users on its own.
- Google DeepMind revealed the latest version of its major large language model. Gemini 1.5 Pro can process text along with video, images, and audio to answer questions.
- Apple is updating its Xcode programming software with AI that can generate code. Apple is currently conducting internal tests of the feature, which could launch to third-party software developers as early as this year.
- Apple researchers debuted "Keyframer," a generative AI tool that animates images using natural language prompts. Users can animate static 2D images based on provided motion descriptions.
- Microsoft and OpenAI reported that hackers, including those backed by Russia, North Korea, Iran, and China, are using generative AI tools like ChatGPT to improve their cyberattacks. The state-sponsored hackers are using AI to refine social engineering, craft phishing emails, translate texts, and debug code.
| | | |
| 7 | | QUICK HITS *This is a sponsored post. | | | |
| Upcoming Events | | MAR
7 | | Portfolio Diversification - Learn from knowledgable experts on the latest trends in alternative investment assets like private equity, crypto, real estate, and the art market. | | | | | | | MAR
27 | | Discover how generative AI can speed up your enterprise, expand your business model, and improve your value proposition. Network with hundreds of AI leaders.* | | | | | | * This is a sponsored event | | | |
 | AI and technology writer | | Beth is a contributing editor and writer of Inside's AI and Tech newsletters. She has written for publications including USA Today, the Arizona Business Gazette, and The Arizona Republic, where she received recognition with a Pulitzer Prize nomination and a First Amendment Award for collaborative reporting on state pension cost increases. You can reach her at Beth.Duckett@yahoo.com. | | This newsletter was edited by Aaron Crutchfield | |
| |
| |
Comments
Post a Comment