AI wants to be free
Imagine you invented one of the most advanced pieces of technology humanity has ever had available. Imagine it cost you thousands of men hours and millions of dollars in computing to develop. Imagine your work was covered in every major newspaper in the world, from The New York Times to Cosmopolitan.
Try to let that idea sink in. You’re about to make tons of money and change the world! Are you excited?
Now, imagine that a competitor comes up with similar technology. They also spent thousands of hours and millions of dollars to build it, but they have a different strategy.
They don’t want to sell it. They want to give it away. For free.
With a single push of a button, someone has completely wiped out your plans to get rich and build a great business. This someone must be completely insane too. Why release such a powerful technology FOR FREE?
It sounds like a completely insane thought experiment, but it’s business as usual in AI. Here are a few stories:
- OpenAI invents GPT-3, an AI model that understands text and can generate prose like a human being. They releases a beta of GPT-3 in July 2020 and make it generally available in November 2021 with a pay-per-use business model
- In July 2022 the Open Science initiative releases BLOOM, a free version of GPT-3
- In August 2022 OpenAI cuts the price of accessing GPT-3 by up to 66%
Here’s another one:
- In April 2022 OpenAI released DALL-E 2, a model capable of generating images of everything you can think of (that’s what I used to make the cover art of this post).
- In August 2022 Stability AI releases a free version of DALL-E 2 called Stable Diffusion.
OpenAI is not the only company getting screwed up. Sometimes, they’re the ones doing the screwing.
- Google, Amazon, IBM, and Microsoft, all sell speech-to-text services. This means you can pay them to transcribe your audio recordings using their AI models.
- In September 2022 OpenAI released Whisper, a free speech-to-text model that performs better than everything else on the market.
- In October 2022 Gianluca Mauro stops paying Google to transcribe his videos and starts using Whisper. OK, this is not historically relevant, but I felt like putting it there.
So is AI a terrible business to be in?
If you’re trying to monetize generic AI models, yes it is. How can you compete when everyone is releasing free stuff? And why are they doing it?
There are three reasons:
- Someone genuinely cares about democratizing access to general AI technology. An example is the Big Science initiative (OpenAI was an example as well).
- Someone isn’t interested in making money from some of those models but sees a nice impact on branding and recruiting by releasing cool-tech-for-free™. This is what Meta does.
- Someone wants to sell you computing power. Google is happy if you have free models, as long as you run them on their cloud and pay your bill.
Why does this matter for you?
If you’re not working for Google or for an AI startup, you’re likely on the side of the table that needs to use AI rather than building fancy algorithms from scratch.
Whatever industry you’re in, the barrier to using AI in your work has never been lower. I’m overwhelmed by the power of the tools available to all of us for little to no money.
The only barrier between you and running AI projects that can disrupt your industry and transform your career is your understanding of AI and your imagination.
Interested in learning how to apply AI to business challenges and adding AI to your skillset? You should join my course AI-ready. Early bird offer expires on Oct 31st.