Google is Behind in AI. Coding is Their Catch-Up Play.

You can almost feel how shocked Google feels to find themselves so behind in AI.

This is the company after all whose search experience has been the starting point for how people interacted with the internet for 20+ years. Holding aside a few notable flops (hello Google+), throughout those two decades Google/Alphabet have never really felt that out of touch or behind.

Until AI.

When it comes to AI, Google has felt like an awkward old guy at a party trying to keep up with the kids to prove they still can. But if they ain't as good as they once was, might they be as good once as they ever was?

On April 16, the NYT broke a story about Google's secretive project Magi, an attempt to reimagine the search experience around a ChatGPT-style bot. The article began "Google’s employees were shocked when they learned in March that the South Korean consumer electronics giant Samsung was considering replacing Google with Microsoft’s Bing as the default search engine on its devices."

...I'm imagining that board room, as Samsung relays the news, the Google employees sit catatonic and slack jawed in disbelief, visions of Steve Ballmer and Bill Gates dancing in their heads...

Google's ChatGPT competitor Bard did them no favors. Rushed out in sheer panic, Bloomberg this week reported how aghast even Google employees were at the product. One called it "cringe-worthy." Another said "Bard is worse than useless: please do not launch." The public's initial reaction suggested those employees had the right of it.

And yet.

On Friday, Google dropped a blog post called "Bard now helps you code" and announced that the AI now had software dev capabilities across 20 programming languages, along with integrations with Google Colab and Google sheets. According to the post it can also review and debug code, as well as translate it from one language to another and explain snippets for those learning.

Could a dedicated focus on AI for coding be where Google catches up? On the one hand, Ben's Bites this morning said that their experiments as well as rumors suggested that Bard was better at coding. On the other, ZDNet tester David Gewirtz wrote "I tested Google Bard's new coding skills. It didn't go well. Given how good ChatGPT is, Google's answer is...embarrassing."

Make matters worse for le Goog, GPT-4-32K is hurdling down the pipeline and one of the most anticipated changes is what it will allow developers to do (see below).

I dunno, man. I'm never counting Google out entirely, but seeing them fall so far behind has been shocking to say the least. And with how lucrative these battles are, I'm particularly worried about ethics and safety corners that get cut to try to catch up.

Today’s Most Important Discussion

Lots of folks talking about GPT-4-32K (including me above!). TL;DR right now GPT-4 has an 8000 token limit - tokens being the building blocks of a language model like words, subwords and characters. That's why sometimes when you try to copy/paste too much text in ChatGPT it can't handle it. With a 4X increase in the token limit, GPT-4 could perform longer text analysis, with improved context retention, better summarization and more extensive creative capacity. Oh, and btw, some viral research is also out that's suggesting handling information up to 2 million tokens is within reach.

I did a video on this today:

"The future rly is now"

Grimes seems pretty serious about this whole "I'll split 50% of royalties with AI generated songs using my voice" thing. Today she started talking about what minimum rules might need to be set and even started talking about automatic royalty distribution through smart contracts.

Thanks for reading! If you want more AI Breakdown:

The AI Breakdown podcast - https://pod.link/1680633614

The AI Breakdown YouTube - https://www.youtube.com/@theaibreakdown

Catch ya tomorrow - NLW