People often ask “What value is in Generative AI nowadays?”
The answer, if you want numbers, is between 43% and -26%.
Key learnings from the BCG study (link in the comments):
1. For the tasks AI is good at, you can expect a double digit increase in productivity.
2. People are not aware yet of what a “task AI is good at” is. We need more testing and transparency.
3. If the task isn’t something AI is good at and you just regurgitate its answers, it actually causes more harm than good.
4. AI assistance is a great leveler of skill. People who want to play catch up with senior talent benefit more from this.
5. Prompt engineering or, in other words, “being better at using chatbots to engage with the LLMs” does help improve performance and avoid worst-case scenarios of using AI.
6. The best uses are when you mix AI & human tasks often. They’re assistants, calculators, not experts.
7. Do not over-rely on AI to grant you creativity.
My own thoughts on some of these:
– Chatbots are also a nice way to train yourself in skills you don’t even know where to start with
– Prepare to fix up AI-generated content, it’s part of the process
– You will take time figuring out the best use cases for it, factor that in
– Do not assume you’ll double or triple your speed, but do assume some speed gains in the long run
– Remember the RISEN framework of prompting.
