AI Voice Hype, Investment, & Regulation

Happy Easter Weekend!

Heading into April, AI’s giving many magical moments- the technological renaissance of the decade.

1/ Amazon’s Mega investment into Anthropic

In fact, Amazon just put in 2.75 billion into Anthropic, its largest investment ever.

Now, why is this significant? Anecdotally and on the LMSYS Chatbot board, Anthropic’s chatbot, Claude-3 Opus, has surpassed GPT-4 for the first time.

Wild- just a couple months ago, GPT-4 was untouchable, and now it looks to be a 2 way race.

GPT-4.5 or GPT-5 is expected to come out in a few months, so be on the lookout….

2/ Hume releases EVI- This stands for empathetic voice interface, and it basically adapts to tonality and speech of the user.

Simply wild.

This is a total game changer- it understands sarcasm, can roast effectively, etc. A great shift forward for natural voice interfaces.

3/ The White house is requiring Chief AI Officers in Government?

In an interesting turn of events, the white house is now requiring all US government agencies to have a chief AI officer.

It is a bit unexpected- the ultimate question is - what will they functionally do? Regulate AI more in ways it doesn’t need to be?

Usually, such is the function of a governmental agency. Of course, in recent history company politics (Google..) seemed to supersede AI effectiveness.

The government entering this arena will likely be much of the same- unless they bring people who actually understand AI at a technological level.

Even then, any innovation the governments can regulate, they will tend toward regulating…

4/ Grok 1.5 on par with GPT-4

It’s improved in reasoning, and now accepts a 128k context window.

Wild how LLMs are starting to catch up, and even surpass GPT-4  in some respects. The LLM race never disappoints.

5/ OpenAI releases a voice generator preview- Voice Engine

Did you think we’d end off on small, peripheral happenings near the end here? Nope, 

OpenAI itself previewed their voice generator on Friday, which can produce 15 second audio samples of natural-sounding speech.

The catch?

They are holding off on releasing it due to potential misuse of the technology.

The AI world always keeps things interesting.