AI’s exponential acceleration is accelerating.
It may have hurt for you to read that sentence (it hurt for me to write it), but it’s true.
@raveeshballah on X posted a chart of the falling cost of intelligence.
As a reminder:
– o1 level intelligence is reasoning intelligence. It’s where the AI can take into account multiple points of view and multiple thought-streams at the same time, in real time, and can show you what it’s thinking about. This is also what DeepSeek is.
– GPT4 level intelligence is what you’re used to if you use ChatGPT on the free plan.
The most outstanding thing to me is not the fact that o1 level intelligence fell 27x in the last 3 months. I think that’s due to DeepSeek’s crazy low pricing model that’s created quite a stir in Silicon Valley.
The most outstanding thing is the cost of GPT4 level intelligence falling 1000x in the last 1.5 years.
When ChatGPT was released in November 2022, it used GPT-3.5. If you used GPT-3.5 today, it would seem like an illiterate toddler compared to what you’re used to.
In March 2023, GPT-4 was released. It was smart but slow and it was expensive. There were rate limits even on paid accounts – call it 30 messages per hour or something like that.
Now, GPT-4 level intelligence is available for free on ChatGPT but it’s also open-sourced through LLaMa and other AI models. Instead of 1 or 2 model vendors with this level of intelligence, there are 8 or 10. Anyone can take the AI and tweak it for their own purpose.
I’m not sure if a technology has ever gotten this cheap this fast.
What does that mean for you?
Build a plan around your business that assumes that these graphs are continuing or accelerating. That might require a change of business model or a change in team structure. It may require bringing on new talent or divesting business units.
In my world, I’m thinking constantly about “what if these trends continue for 18 months?” “What if they continue for 5 years?”
In our business, we do 3 things:
1. Coach teams on how to use off the shelf AI tools to save 5+ hours per week
2. Perform Capabilities Workshops (process reviews) to determine what processes in your business are best for automation and will free up your team the most.
3. Build AI/automations to automate those processes.
As AI continues to get better, how could we dial up our coaching so everyone can save 10+ hours per week with off the shelf AI tools?
As AI gets better, what’s the best structure for a development team? How do our devs leverage the AI instead of getting replaced by it?
How could AI help in the Capabilities Workshop process to free up our team from mundane documentation and also help ensure that no details get missed?
As the AIs get better, how could we deliver software 2x or 3x as fast?
As we build out our own strategy, we’re assuming these trends continue or get faster.
I recommend you do the same.
I’m happy to help if you’re curious about what we’re thinking or how it might apply to you. Just reply to this email.