Welcome to our weekly roundup of handcrafted AI news.
This week, OpenAI showed off more vaporware we can look forward to.
You may not need a data center to train your model.
And big tech companies are writing anti-establishment letters to ‘save’ AI.
Let’s get started.
Strawberry rumor
You may have noticed strawberries appearing disproportionately in AI discussions this week. AI models still can’t tell how many Rs are in the word “strawberry,” and Sam Altman randomly posted pictures of strawberries on X.
Why is the AI world buzzing about strawberries? In a frustratingly consistent fashion, OpenAI has been leaking rumors and gossip about a new model called Strawberry and another called Orion.
Rumor has it that they’ve shown this to the federal government, but who knows if it’ll ever be put to practical use, or if it’ll end up in the “too risky” category alongside the GPT-4o voice assistant.
ChatGPT Users will also have to wait a little longer to use the following features: SoraVoice and SearchGPT… pic.twitter.com/WRlQDEVZKb
— Ashutosh Shrivastava (@ai_for_success) August 28, 2024
The first experience with voice interaction via ChatGPT may actually come from Apple, with the iPhone 16 and apologize Siri powered by ChatGPT will soon have a significantly improved intelligence release schedule.
Luckily, Google is rolling out real products, not just AI steamroller software. The company unveiled three new experimental products. Gemini Among the models featured this week are some smaller models that punch well above their weight.
Game Changer
Do you need a huge, expensive data center full of NVIDIA hardware to train AI models? Maybe not. An AI training optimization algorithm called DisTrO could completely change the way AI models are trained.
Imagine the computing power of internet-connected PCs and GPUs around the world training a model similar to how Bitcoin mining works. DisTrO can make it happen.
AI in the workplace
Amazon Web Services (AWS) CEO Matt Garman told employees that AI could replace developers within the next 24 months, a reality that has sparked interesting discussions about changes in work streams and the value of creativity.
The Chinese are also fans of AWS. With a little creativity and AWS, they were able to access NVIDIA chips that were on the US export ban list.
While Garman and others are trying to reassure their human workers, Swedish FinTech company Klarna is embracing AI as it plans to reduce its headcount by 50%. Klarna’s CEO says AI will help it double its sales per employee.
Will Laws Hinder AI?
Several AI safety regulations have recently come into effect or are about to become law, but they have little support.
OpenAI has been critical of California’s AI safety bill, which could become law in the coming days. A former OpenAI employee has written a damning review of it. OpenAIHe spoke out against the AI safety bill and called Sam Altman a “liar” for supporting AI safety.
Is SB 1047 a bad idea, or is Sam Altman just a crybaby?
https://t.co/BI85vrqAaT pic.twitter.com/TGphg58bRP
— Cicero ⏸️ (@PauseusMaximus) August 26, 2024
Meta CEO Mark Zuckerberg and Spotify CEO Daniel Ek are not fans of the EU AI Act. They wrote a letter to EU regulators criticizing the regulations that came into effect this month. The short version of the letter reads: ‘AI can do great things for us, but you are preventing it from happening.’
In other news…
Here are some other clickable AI stories we enjoyed this week:
This is crazy. The model is generating this game in real time while the user is playing.
If we mapped the progress/trajectory of most other AI modes here, we would be generating AAA quality games in real time within a few years. pic.twitter.com/RQi8396j8u
— Matt Schumer (@mattshumer_) August 28, 2024
It’s over now.
Do you think we’ll be able to try Strawberry or Orion any time soon? I keep telling myself that I’m done waiting for GPT-5, but who am I fooling?
The DisTrO story could be huge. Would you connect your PC to be part of a ‘hive mind’ to train a huge open source model? If you get decentralized AI, it will be difficult to enforce AI alignment and safeguards.
We’ll be watching California’s SB 1047 bill closely over the next few days. If it passes, we could see a lot of AI experts joining Elon Musk when he heads to Texas. Or will OpenAI and Meta just sit tight?
Let us know what you think, follow us on X, and send us links to interesting AI research or news we might have missed.