- The Daily Update
- Posts
- Senators question OpenAI's safety
Senators question OpenAI's safety
AND Microsoft launches Bing generative search
Jack Shields
July 25, 2024
Welcome back!
The LLM race is heating up again. Paris-based Mistral has already responded to Meta’s Llama 3 launch. That and more…
In today’s Daily Update:
🗞️ US Senators demand details on OpenAI’s safety efforts
🤖 Microsoft redesigns Bing’s search experience
📸 Mistral Large 2 arrives one day after Llama 3
🚨 AI Roundup: Four quick hits
Read time: 2 minutes
TOP STORY
🗞️ US Senators demand details on OpenAI’s safety efforts
DALL-E 3
A group of U.S. Senators sent a letter to OpenAI CEO Sam Altman demanding information about the company’s safety practices by August 13.
The rundown:
The letter is a response to recent reports of potential safety risks at OpenAI.
Among other concerns, OpenAI is facing scrutiny from former employees for prioritizing product launches over safety.
The company reportedly pushed its latest model (GPT-4o) through safety testing in just one week.
OpenAI previously made safety commitments to the Biden Administration, and is currently working with U.S. agencies on AI-based cybersecurity tools.
Why it matters: As we saw a couple weeks ago, OpenAI is far from reaching the pinnacle of AI. The issue is that safety shortcomings that go unchecked now will likely persist as the company climbs the scale toward AGI. Holding OpenAI accountable for recent safety lapses is probably in everyone’s best interest.
AI TOOL OF THE DAY
🤖 Microsoft redesigns Bing’s search experience
Source: Microsoft Bing Blogs
Microsoft Bing is experimenting with a new layout that relies on AI-generated answers instead of traditional search results.
Key points:
The new layout pushes standard search results (web pages) to a sidebar, with the core of the page featuring AI-generated summaries that answer user queries.
Microsoft is currently rolling out Bing generative search to a “small percentage” of users.
Google has faced criticism for a similar feature called “AI Overviews,” which delivers inaccurate answers to certain queries fairly frequently.
The future of search: Generative AI will more than likely upend the traditional search experience, forcing companies like Google and Microsoft to frantically explore ways for their search engines to match the convenience of chatbots. This is good news for consumers, as we’re set to see the emergence of AI search engines that fulfill our queries more thoroughly.
BUSINESS SPOTLIGHT
📸 Mistral Large 2 arrives one day after Llama 3
DALL-E 3
Mistral AI just announced Mistral Large 2, the latest version of its flagship model that rivals Meta’s newly released Llama 3.
The details:
Mistral Large 2 is a 123 billion parameter model that offers improved capabilities in code generation, math and reasoning.
The model appears to outperform Llama 3 across most benchmarks, and even competes with leading commercial models like OpenAI’s GPT-4o.
Mistral Large 2’s 128k context window allows it to process roughly 100,000 words or 300 pages in a single interaction.
The model also supports dozens of languages and over 80 coding languages.
The big picture: OpenAI believes it is on the verge of reaching the second level on its scale to AGI. Based on the benchmarks of new models from several other companies, it looks like there are a few players preparing to make the same leap. Expect to see a major breakthrough in language model capabilities by the end of the year.
MORE TRENDING NEWS
🚨 AI Roundup: Four quick hits
Source: Michael Zagaris / Getty Images
OpenAI develops Rule-Based Rewards (RBRs) method that improves model safety behavior (possible response to today’s top story).
Chinese video-generating model appears to censor politically sensitive topics.
Former NFL quarterback launches new AI storytelling platform to help creators own their stories.
U.S. Senate passes a bill cracking down on sexually explicit deepfakes.
THAT’S A WRAP