Innovating with AI - Did AI really save 4,500 years of work?

Hey Innovator,

Over in the Innovating with AI student Slack, we've been talking a lot about big stats from large companies implementing AI:

• Amazon says AI coding assistants have already saved "4,500 developer-years" and $260 million
  ​
• Goldman Sachs' CIO joined the Odd Lots podcast to talk about how they're building hybrid LLMs that their team can use securely
  ​
• ​Walmart is using AI to write and edit its product catalog content, getting the work done with 1% of the people it would have needed otherwise.

Now, most of the students in our Incubator and Consultancy programs don't work for these gigantic companies – they have much leaner businesses with better work-life balance.

However, looking at these giant firms can show us what happens when you multiply AI's efficiencies times thousands of people, and it's pretty crazy.

Let's dig into each one.

At Amazon, coding assistants are making software upgrades much faster. This means that a job that would take a human developer 50 days can be done in 1 or 2 days instead. (And this is maintenance "grunt work" that coders don't enjoy anyway, the least creative type of work.)

So basically the AI is doing the software equivalent of an oil change, and doing it almost instantaneously. Multiply that across a giant company and you end up saving 9,000,000 hours (nine million hours!), or the equivalent of a full-time workload x 1 year x 4,500 people.

•

Goldman has built "cordoned off" LLMs to allow its employees to run sensitive data through AI without risking confidentiality or security. They actually ended up not building their own large language model (which costs $100s of millions of dollars to do from scratch). Instead, they basically built an internal platform that totally locks external eyes out of their proprietary data, while still being able to use OpenAI and Google's advanced large language models.

Since Goldman is huge, they can get first dibs to this special type of data privacy deal. But, I think this is a sign of what's to come for normal companies and consumers in a few years.

•

Walmart barely needs copywriters anymore. Instead, it is using LLMs to write the content that appears in its product catalogs. They also appear to be using image generators or upscalers (making a low-res image look more clear and crisp at large sizes) throughout their catalog. Walmart's CEO says:

This one is actually pretty low-hanging fruit for any e-commerce company with a large catalog, and something we've already started exploring with our software development clients. Walmart is just doing it at tremendous scale, having used AI to updated 850,000,000+ data points so far.

•

What does this mean for consultants and creators?

First, it means that every AI idea in your head is going to get more viable as big companies adopt AI more aggressively in the next few years.

If you're building an AI consultancy, small and medium businesses are going to be clamoring for you to help them do a miniaturized version of the cool stuff they see the giant corporations doing.

And if you're building an AI product or process, you're gonna have access to better and better tools every few months.

These big companies talking publicly about how they're saving hundreds of millions of dollars changes pretty much everything about the economy. Especially when you compare it to the Internet and computers in general, which, while transformative, have (somewhat bafflingly) never made a dent in measurable productivity.

To put it another way, AI is on track to be a bigger deal than computers.

Have a great weekend,

– Rob at Innovating with AI