Wake Up.
I was working on my Table Driven AI system, where I record my evaluation rules for documents (like resumes) and multi-stage instructions, prompts at a time, so I can get robust, longer responses…And I was running up the AI bill. Even with caching turned on with the Anthropic API (Claude), doing a thorough evaluation of a resume and enhancing the data with alternative skill names and generic terms for the sake of semantic search, the cost was hovering at about 50 cents.
I decided to switch to a newer model just released, Claude 3.5 Haiku, and the cost went down to 9 cents! With no loss of quality. What will the cost be in six months? 3 cents? 1 cent? Free?
People talk about the cost of intelligence falling. We see smaller LLMs, purpose built LLMs, and tooling, like LlamaIndex and LangChain, making intelligent automation so affordable and easier to implement that it makes more complex and intense operations feasible. You remove cost from the equation, you’re left with these barriers to entry and proficiency:
- Knowledge — Acquisition of knowledge, and hence ability, comes from learning and doing. And doing that takes practice. You have to find the time
- Opportunity — You will not find opportunity just waiting to be discovered. You have to make your own opportunity. Why? The majority of people don’t know what’s possible now. The majority of companies do not have initiatives to initiate the most drastic transformation they are ever to undertake. For myself, I had to feel the itch and identify the benefit for using AI — that was reading resumes — and identify the mission: Help people find the talent and make the talent discoverable.
- Confidence — I’ve listed the technological hurdles and the challenge of finding a problem to solve, the other challenge is this: It’s going to take courage to step into an area so new, so quickly advancing, that you’re effectively a learner — a fresher, as I hear some people refer to those just starting off — once again just like you were when you didn’t know anything and didn’t even know you didn’t know anything. For a lot of people, that’s an uncomfortable place.
- Time — But this is the easy one. Make the time, find the time, prioritize the time.
In the news and on LinkedIn, people talk about the impact of AI on jobs.
For me, the Amazon Return To Office is about this: We do not need the skills of the talent that we hired. The talent that feels that they can go to where they’re more appreciated can, maybe will, find opportunities that reward them for the prestige that they acquired and their achievements, but Amazon does not need them. Why? Because what Amazon needs is not the most senior people in the org that once wrote great code and excellent algorithms. Have you seen the AWS list of product offerings? No one needs those many services. Instead what Amazon needs are solution architects that can be matched to clients and expand partnerships, and with AI, they can do that.
With AI, our proper role is, in fact, that of the architect, for now. In the future, it will be that of a product designer/manager. In the even further out future, it will be that of the Imagineer. When you can do anything, the only thing that matters is what you want, and the only thing limiting you in what you want is having the right words to express it.
For now, though, our responsibility is to learn what we can, as quickly as we can, and build solutions that have value and that we can capitalize on. Our goal should be to keep ourselves relevant and contributing in a future that is already here but is yet to be fully recognized.
Because people are beginning to wake up and they’re beginning to see what’s possible.