Welcome to our weekly roundup of human-crafted AI news.
This week OpenAI handed out API goodies.
California’s AI safety bill got killed.
And NVIDIA surprised us with a strong open model.
Let’s dig in.
Here come the agents
OpenAI didn’t announce any latest models (or Sora) at its Dev Day event, but developers were excited over latest API features. The Realtime API might be a game-changer for making smarter applications that talk with users and even act as agents.
The demo was really cool.
There have been rumors about OpenAI going the “for-profit” route and awarding Sam Altman billions of dollars in equity but Altman dismissed these. Even so, the corporate is on the drive for more investment they usually’ll expect some return for his or her money.
Apple has integrated OpenAI’s models into its devices but dropped out of the newest funding round for OpenAI which is predicted to lift roughly $6.5 billion.
We’re unsure why Apple doesn’t need a piece of the OpenAI pie however it may need something to do with latest developments with its Apple Intelligence. Or possibly it’s related to Sam Altman’s demand for exclusivity.
Tell me you will have no moat without telling me you will have no moat. pic.twitter.com/3I18MosvOg
Kill bill
Gavin Newsom had to determine between putting a security rev limiter on AI developers or letting them go full steam ahead. In the top, he decided to veto California’s SB 1047 AI safety bill and offered some interesting explanation why.
Are we actually at some extent where we face real AI risks yet?
Well that escalated quickly https://t.co/xhZCITRJjE pic.twitter.com/aLZn4blS8G
Newsom has signed a variety of AI bills during the last month related to deepfakes, AI watermarking, child safety, performers’ AI rights, and election misinformation. Last week he signed AB 2013 which can really shake things up for LLM creators.
The bill says that on or before January 1, 2026, developers will need to offer a high-level summary of the training dataset of any models comprised of January 1, 2022 onward if the model is made available in California. Some of those requirements could air some dirty secrets.
More EU AI regs
The EU is clearly quite a bit more concerned about AI safety than the remainder of the world. Either that or they simply enjoy writing and passing laws. This week they kickstarted a project to write down an AI code of practice to try and balance innovation & safety.
When you see who heads up the security technical group you’ll have an excellent idea of which way they’ll be leaning.
Liquid foundation models
Transformer models are what gave us ChatGPT but there’s been numerous debate recently about whether or not they might be as much as delivering the subsequent leap in AI. An organization called Liquid AI is shaking things up with its Liquid Foundation Models (LFMs).
These aren’t your typical generative AIs. LFMs are specifically optimized to administer longer-context data, making them ideal for tasks must handle sequential data like text, audio, or video.
The LFMs achieve impressive performance with a much smaller model, less memory, and fewer compute.
NVIDIA opens up
Nvidia just dropped a game-changer: an open-source AI model that goes head-to-head with big players like OpenAI and Google. Their latest NVLM 1.0 lineup, led by the flagship 72B parameter NVLM-D-72B, shines in each vision and language tasks while also leveling up text-only capabilities.
With open weights and NVIDIA’s promise to release the code, it’s getting harder to justify paying for proprietary models for numerous use cases.
NVLM-D benchmarks. Source: arXiv
Just say know
A brand new study found that the newest large language models (LLMs) are less more likely to admit after they don’t know a solution to a user’s query. When users ask these models a matter they’re more more likely to make something up moderately than admit they don’t know the reply.
The study highlights the necessity for a fundamental shift within the design and development of general-purpose AI, especially when it’s utilized in high-stakes areas. Researchers are still trying to grasp why AI models are so keen to please us as an alternative of claiming: ‘Sorry, I don’t know the reply.’
AI inside
It looks as if everyone seems to be slapping an “AI” label on their products to drag in customers. Here are a couple of AI-powered tools which can be actually value trying out.
Bluedot: Record, transcribe, and summarize your meetings with AI-generated notes with out a bot.
Guidde: Guidde magically turns your workflows into step-by-step video guides, complete with AI-generated voiceovers and pro-level visuals, all in only a couple of clicks.
In other news…
Here are another clickworthy AI stories we enjoyed this week:
And that’s a wrap.
If you reside in California we’d like to know the way you are feeling about SB 1047 getting vetoed. Is it a missed opportunity for AI safety or a positive step that may see us get AGI soon? With powerful open-source models like NVIDIA’s latest bombshell, it’s going to be harder to control LLMs anyway.
OpenAI’s Realtime API was the highlight this week. Even in the event you’re not a developer, the prospect of interacting with smarter customer support bots that talk over with you is pretty cool. Unless you’re employed as a customer support agent and also you’d prefer to keep your job that’s.
Let us know what you think that, follow us on X, and send us links to chill AI stuff we could have missed.
