| |
| Here's a look at today's AI briefing: - OpenAI announces AI voice cloning tool but won't release it publicly.
- Microsoft and OpenAI partner on $100B data center project.
- Hillary Clinton warns that AI threatens this year's global elections.
- Elon Musk's AI company announces latest large language model.
- New York City's AI chatbot generating illegal, harmful advice.
Beth
p/beth-duckett | |
| 1 | | OpenAI previewed Voice Engine, a voice cloning technology that can mimic speakers based on 15-second audio samples, but is withholding a public release over safety concerns. "We recognize that generating speech that resembles people's voices has serious risks, which are especially top of mind in an election year," the company said in a blog post. More: - OpenAI first developed Voice Engine two years ago and has been testing it with a small group of partners.
- Voice Engine, based on the company's text-to-speech API, generates "natural-sounding speech" with "emotive and realistic voices" that closely resemble the original speaker.
- The tool can recreate voices in languages including English, Spanish, French, or Chinese.
- Potential applications include reading support, language translation, and helping non-verbal individuals and those with speech impairments regain their voices.
Zoom out: - On its blog, OpenAI shared early demos and samples of Voice Engine.
- The company scaled back the tool's rollout based on feedback from industry experts, policymakers, educators, and creatives.
- It will decide whether to have a wider rollout after more testing and discussions.
|     | |
| 2 | | Microsoft and OpenAI are partnering on an over $100B U.S. data center project that would house an AI supercomputer called "Stargate," according to sources cited by The Information. The supercomputer, made up of millions of AI chips, could launch as early as 2028, More: - Microsoft is expected to finance the project, which could cost upwards of $115B, according to sources who saw Microsoft's cost estimates and spoke with OpenAI CEO Sam Altman.
- At that cost, the project would be around 100 times more expensive than some of the largest data centers today.
- Stargate's power requirements, estimated to be "at least several gigawatts," could force Microsoft and OpenAI to explore alternative energy sources like nuclear power, the report added.
Zoom out: - Stargate is the largest in a series of supercomputers planned by the companies over six years.
- As the fifth phase in the series, Stargate would be preceded by a smaller fourth-phase supercomputer for OpenAI, which is expected to come online around 2026.
| | | |
| A message from our sponsor, Top Class Actions. |  | | Analyzing the Impact of Visa-Mastercard's $5 Billion Settlement! Attention all business owners and those in financial authority in making business financial decisions, your portion of a $5.6 Billion settlement is waiting to be claimed. The deadline to submit your claim is rapidly approaching, and, if you miss it, your money goes to other claimants. Stay ahead of the curve and leverage this settlement to drive innovation and growth within your organization. Are you: - A business owner or make financial decisions for a business
- Accepted Visa or Mastercard from Jan. 1, 2004 to Jan. 25, 2019
- Looking to claim your part of the settlement
Learn More | |
|
| 3 | | Hillary Clinton cautioned that foreign governments using AI to spread disinformation could threaten this year's elections. During a talk Thursday at Columbia University, the former senator and secretary of state said 2024 "is the year of the biggest elections around the world since the rise of AI technologies like ChatGPT." What happened: During the talk, Clinton and other speakers said government and tech companies need to team up to combat disinformation. Election officials expressed concern about AI-generated deepfake videos that could discredit campaigns and manipulate people at a mass scale. Quotable: Clinton spoke about how Russia-backed campaigns spread disinformation about her on social media during her 2016 presidential run, adding that the effort was not fully understood. But that campaign "was primitive" compared to the "leap in technology" including AI since then, Clinton said. "Defamatory videos about you is no fun — I can tell you that," she added. "But having them in a way that … you have no idea whether it's true or not. That is of a totally different level of threat." Zoom out: This year's election cycle already saw the use of AI through a robocall mimicking President Joe Biden's voice, which urged New Hampshire voters not to cast their ballots in the primary. Shortly after, the FCC voted to ban the use of AI-generated voices in scam robocalls. Now, New Hampshire's state House is advancing a bill requiring political ads using deceptive AI to disclose its use, making it the latest state to take action against AI's threats to elections. By the numbers: According to a Pew Research Center survey, 38% of Americans distrust information from ChatGPT about the 2024 U.S. presidential election, with only 2% expressing high levels of trust. Distrust spans both political parties, with around 40% of Republicans and Democrats lacking trust in ChatGPT's election responses. Despite this, very few Americans have actually used the chatbot for election information, with only 2% reporting that usage across party lines. 
| | | |
| 4 | | Elon Musk's AI startup unveiled its newest generative AI model, Grok-1.5, with improved mathematics and reasoning capabilities. xAI will release the Grok-1.5-powered chatbot to early testers on Musk's social media platform X next week. More: - Grok-1.5 outperforms Grok-1, scoring over twice as high on the MATH benchmark and more than 10 percentage points higher on the HumanEval test for programming and problem-solving.
- Grok 1.5 also surpassed competitors like Claude 2, Gemini Pro 1.5, and GPT-4 on some problem-solving benchmarks.
- However, Grok 1.5 scored 50.6% on the high school math test, lower than Claude's 61%, Gemini's 58.5%, and GPT-4's 52.9%.
- On the GSM8K grade school test, Grok 1.5 scored a solid 90% but still fell short of Anthropic, Google, and OpenAI.
Zoom out: - Grok-1.5 can process contexts of up to 128,000 tokens, allowing it to handle information from much longer documents.
- In a post on X, Musk noted that Grok 2 is now in training and "should exceed current AI on all metrics."
| | | |
| A message from our sponsor, Dun & Bradstreet. |  | | See Scores and Ratings Insights—and Fewer Surprises Stay informed about your business's credit by regularly monitoring changes to your business scores and ratings, which include: - PAYDEX(R) Score
- Delinquency Score
- Failure Score
- Maximum Credit Recommendation
Understanding the scores and ratings in your business credit file may provide insights into your business’s financial strengths and weaknesses that can help you make important business decisions. Try it for free | |
|
| 5 | | New York City's AI chatbot is generating inaccurate and potentially illegal advice, according to The Markup. The MyCity bot has told business owners that they can take employees' tips and other incorrect information. More: - The AI bot, powered by Microsoft Azure AI, was designed to help NYC residents with housing and small business concerns.
- However, the Markup and Quartz found that the chatbot sometimes gives out false information that can violate housing and worker rights laws.
- When asked if a manager can take a portion of their employees' tips, the chatbot reponses "Yes," erronesouly citing NYC's Payroll and Tip Reporting information.
- It also stated that landlords are not required to accept tenants on rental assistance. According to NYC's website, discriminating based on lawful sources of income, including assistance, has been illegal since 2008.
Response: - The chatbot pilot program has provided thousands with accurate answers while warning them of potential risks, a spokesperson for the NYC Office of Technology and Innovation said.
- The city "will continue to focus on upgrading this tool so that we can better support small businesses," she added.
| | | |
| 6 | | Weekly roundup — The top AI stories you might have missed this week: - The White House announced new rules for U.S. government agencies to make sure their AI tools don't harm the public. Over the coming months, federal agencies will have to appoint chief AI officers, report their AI use, and set up safeguards against risks.
- Amazon said it will invest a further $2.75B into Anthropic, adding to its initial $1.25B investment in the AI startup from last September. The latest tranche completes Amazon's pledge to invest up to $4B in Anthropic, a competitor to OpenAI and Google known for its Claude models.
- Google is expanding its AI-assisted search to a small subset of U.S. users as it moves toward a more widespread public rollout. The AI search feature, called Search Generative Experience (SGE), was previously only available to people who signed up to test it.
- Apple is expected to unveil new generative AI features for iPhones and more during its Worldwide Developers Conference (WWDC) in June. WWDC kicks off on June 10 with a keynote announcing updates to its operating systems, including iOS, iPadOS, and macOS.
- Apple chose Baidu's generative AI to power upcoming features on its iPhones and Macs in China, according to local media. Apple's Ernie Bot is expected to appear on the upcoming iPhone 16 this fall, as well as future versions of Mac and iPhone operating systems.
- Stability AI CEO Emad Mostaque resigned last weekend, days after several key researchers behind its technology announced their departures. The AI startup named COO Shan Shan Wong and CTO Christian Laforte as interim co-CEOs while searching for a replacement.
| | | |
| 7 | | QUICK HITS - Want to reduce your AWS costs? Get your AWS cloud spend under control and maximize your AWS cost savings today.*
- The U.S. House of Representatives has banned congressional staffers from using Microsoft's Copilot assistant The House previously banned the free version of ChatGPT last year.
- DeepMind CEO Demis Hassabis received a knighthood in the U.K. for his contributions to AI. Hassabis co-founded the AI research lab DeepMind in London in 2010, before it was acquired by Google in 2014.
- An unnamed BBC project has replaced British actress Sara Poyzer with an artificially generated voice. The move comes after BBC director general Tim Davie said the company would "proactively deploy" AI.
- Elevate your beauty brand on Amazon. Pattern's guide unveils this multi-billion dollar market.*
*This is a sponsored post. | | | |
 | 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 Beth Duckett | |
| |
 | Top Class Actions has helped law firms across the country successfully find plaintiffs for class action lawsuits & mass torts since 2008. | |
|
| |
|
| |
Comments
Post a Comment