| |
| Here's a look at today's top AI stories: - Samsung unveils Galaxy S24 smartphones powered by Google AI.
- DeepMind AI can solve geometry problems at the Olympiad level.
- U.S. lawmakers propose a federal ban on nonconsensual deepfake sharing.
- U.N. chief calls on world leaders to do more to tackle AI.
Beth
p/beth-duckett | |
| 1 | | Samsung today revealed its latest flagship Galaxy S24 smartphones, which are equipped with AI tools powered by Google's Gemini models. The AI features include photo editing, live phone call translation, voice recording transcription, web article summarization, and more. More: - The lineup consists of the standard $800 Galaxy S24, the $1,000 Galaxy S24 Plus, and the $1,300 Galaxy S24 Ultra.
- All of the phones run on Qualcomm's Snapdragon 8 Gen 3 applications processor, which can handle some of the AI functions like live translation.
- Users can toggle between two AI models: Gemini Nano on the device itself, and a more powerful Pro version in the cloud.
- Some AI features are Samsung's, and others were co-developed with Google.
AI Features: - Live translation of phone calls.
- Chat Assist for text message tone adjustment, translation, and spell-checking.
- Text summarization in the Notes app.
- Generative edits for manipulating objects in photos.
- "Circle to Search," letting users search for objects in images by circling them first.
Zoom out: |     | |
| 2 | | Google DeepMind has created an AI system that can solve geometry problems at the Olympiad level. The "AlphaGeometry" system solved as many geometry problems as an average human International Mathematical Olympiad (IMO) gold medalist, who is typically a high-school mathematician. More: - Geometry poses a challenge for AI researchers due to its need for both creativity and rigidity.
- It also lacks the extensive data available to text-based AI models trained on the web.
- To get around this, AlphaGeometry combines a neural language model with a "symbolic deduction engine" to validate the former's ideas using mathematical rules.
- To train the model, DeepMind generated 100 million synthetic theorems and proofs.
Results: - AlphaGeometry was then tested on 30 challenging IMO geometry problems.
- The system successfully solved 25 of the problems within the 4.5-hour time limit, surpassing the previous best system that solved only 10.
- The results were published in a study in Nature this week. DeepMind has also open-sourced the code for AlphaGeometry.
| | | |
| A message from our sponsor, Faros AI. |  | | In scaling a tech company, developing an engineering vision and execution strategy becomes a cornerstone for success. Get the field-proven playbook Mustafa Furniturewala used as SVP of Engineering at Coursera (NYSE: COUR) to 10x his team. With this playbook, you’ll learn how to: - Craft a three-year vision aligned with company goals utilizing four critical inputs
- Develop an execution strategy to deliver the vision with a 20/80 capacity split
- Utilize a 2x2 resource allocation framework that helps prioritize unplanned requests
- Identify leading indicators of team health and architectural and organizational “drift”
Get your copy of the Engineering Leadership Playbook, compliments of the engineering intelligence experts at Faros AI, today. Get the Playbook | |
|
| 3 | | U.S. Rep. Joseph Morelle (D-N.Y.) has reintroduced a bill proposing a federal ban on the sharing of nonconsensual sexually explicit deepfakes. The "Preventing Deepfakes of Intimate Images Act" now includes a co-sponsor, U.S. Rep. Tom Kean (R-N.J.). More: - In addition to criminalizing nonconsensual "deepfake" sharing, Morelle and Kean's bill would permit victims to file civil lawsuits against offenders.
- In a statement, Motelle said the legislation would "finally make this dangerous practice illegal and hold perpetrators accountable."
- The bill is a response to an incident at a New Jersey high school where male students shared AI-generated explicit images of female classmates.
- Francesca Mani, one of the victims, addressed a news conference about the bill, saying, "Deepfakes are happening every single day to women everywhere."
- Kean also introduced the AI Labeling Act in November. If passed, it would force AI developers to include visible disclosures on products created with their AI tools.
Zoom out: - Computer science professors now believe that AI's deepfake capabilities are at the advanced level, according to an Axios-Generation Lab-Syracuse University AI Experts Survey.
- Respondents chose digital authenticity certification as the most effective strategy to combat AI deepfakes, followed by public education, enhanced detection technology, and stricter regulations.
| | | |
| 4 | | U.N. Chief António Guterres has called on world leaders to take more action to tackle the twin threats of AI and climate change. During Wednesday's World Economic Forum, Guterres pointed to the lack of effective global strategies to deal with the threats, blaming "geopolitical divides." More: - In a special address, Guterres said that each interaction with generative AI raises the threat "of serious unintended consequences."
- The U.N. secretary-general said the world appears "powerless to act together" in the face of the existential threats of climate change and "runaway development of artificial intelligence without guardrails."
- He argued that powerful tech firms are already pursuing profits with a "reckless disregard" for social impact, personal privacy, and human rights.
- While AI has major possibilities for sustainable development, Guterres also warned that AI is likely to exacerbate inequality, citing a new IMF report.
- That report warned that AI could disrupt nearly 40% of the world's jobs, especially in high-income countries, further widening the wealth gap.
Solutions: - Guterres urged governments to collaborate with tech companies to create AI risk management frameworks and future harm prevention.
- He also called for a "systematic effort" to open up AI access for developing economies to harness its "enormous potential."
| | | |
| A message from our sponsor, Buzzboard. |  | | Grow Your Sales to Local Business! If you sell to small, local and mid-sized businesses, you need our AI-powered digital sales platform! We fuze our one-of-a-kind exhaustive database on SMBs with generative AI to enable hyper-personalized sales cadences that work. With over 6,000 signals on over 30 million businesses, you’ll never have to resort to spammy templates again. BuzzBoard Ignite makes selling to SMBs more impactful. Our platform uses your product data, artificial intelligence, and our database to create bespoke prospecting e-mails, conversation starters and more, to open new doors, win more customers and close bigger deals. Join sales leaders like yourself who have already discovered how tools like our SWOT analysis and competitive benchmarking reports build more confidence before going into sales calls.
Sign up for a free trial today! Free Trial - No Credit Card Required | |
|
| 5 | | Security researchers discovered a vulnerability in some GPUs that could allow local attackers to access sensitive AI data, including a user's chatbot sessions. According to MacRumors, the flaw is especially worrying as GPUs are increasingly used to process data for large language models. More: - Cybersecurity firm Trail of Bits first uncovered the security flaw, dubbed "LeftoverLocals."
- The vulnerability was found in GPUs made by Apple, AMD, Qualcomm, and Imagination.
- Attackers with local access to a phone or other device can access and read data, like ChatGPT sessions, that was processed in the GPU's local memory.
- Apple has since patched the security flaw in newer devices using A17 Pro and M3 chips.
- However, older models like the iPhone 12 and M2 MacBook Air are still susceptible.
| | | |
| 6 | | 🍋 FRESH FUNDING 🍋 - Norwegian AI and robotics startup 1X, formerly known as Halodi Robotics, closed its Series B round at $100M. EQT Ventures led the round with additional backing from Sandwater, Skagerak Capital, Samsung NEXT, and Nistad Group.
- Smart hospital platform Artisight Inc. raised $42M in Series B funding from investors, including chipmaker NVIDIA.
- Japanese AI startup Sakana AI raised $30M in seed funding at a valuation of $200M. Lux Capital led the round, with additional participation from Khosla Ventures, Sony Group Corp., and NTT. The Tokyo-based AI startup, founded by ex-Google employees, is developing smaller AI models.
- El Salvador-based AI startup Spot Technologies secured $2M in funding from investors including Femsa Ventures, Bridge Latam, and Daedalus. Spot's VisionX product, used by Walmart and other retailers, leverages AI to analyze consumer and theft patterns.
For the latest on AI startup funding and more in the VC community, subscribe to our Inside Venture Capital newsletter. | | | |
| 7 | | QUICK HITS - Are you creating the unique user journeys your customers demand? You can by Mastering Personalization with AI.*
- The FDA has approved DermaSensor's handheld AI device for non-invasive skin cancer detection. Doctors can use the device to analyze suspicious skin lesions for basal cell carcinoma, squamous cell carcinoma, and melanoma.
- Google announced AI-powered multisearch for mobile phones, including the Pixel 8 and new Samsung Galaxy S24 models. The feature lets U.S. users search using both text and an image simultaneously.
- TikTok owner ByteDance quietly released four new generative AI apps powered by OpenAI's GPT. Three of the apps, Cici AI, ChitChop, and Coze, let users create their own chatbots, while a fourth app called BagelBell generates stories that adapt to reader choices.
*This is a sponsored post. | | | |
| Upcoming Events | | MAR
7 | | Alternative investment assets to know in 2024. | | | | | | * 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 notable publications like USA Today and the Arizona Business Gazette. During her time as a public policy reporter at The Arizona Republic, she received recognition with a Pulitzer Prize nomination and a First Amendment Award for her collaborative reporting on pension cost increases. Beth also authored a book on the solar photovoltaic industry in 2016. You can reach her at Beth.Duckett@yahoo.com. | | This newsletter was edited by Aaron Crutchfield | |
| |
 | Faros AI improves engineering productivity and the developer experience with centralized visibility and AI insights. | |
|
| |
|
| |
Comments
Post a Comment