We know AI can do a lot of things, but not many things well (yet). Nevertheless, people including me are using more and more AI, so it is useful to think about what it can do and do well in the real world.
Recently I read a WSJ article that summarizes reasonably well the status quo. It points out the following categories:
• Call centers and agents. This is mostly AI bots, either speaking or writing, to answer customers questions. I agree this is going to be very useful.
• Medical application. Mostly recording keeping. But “this could be the first step toward automated generative diagnostics”. Absolutely. With enough data, AI should be able to help with diagnosis and treatment. Very promising. Could be huge.
• Legal. “Several legal AI tools have hit the market to augment lawyers as they peruse databases of contracts and legal code.” Agreed. AI will be very good at researching and synthesizing information. Considering the legal profession is pretty large (1.3% of GDP), this is quite noticeable. Whether it can do arguing or cross examining, or even generate memory quotes such as “If it does not fit, you must acquit”, we will see. Maybe it will.
• Companions. This is a real one. Since the dawn of human civilization, we have been seeking love/lust and companionship, virtual or not. And AI can do a fairly good job in this area, I think.
• Coders. Google claims that more than 25% of its internal source code is now AI-generated. Salesforce has announced a hiring freeze for software engineers. Facebook hopes to automate “midlevel” software engineers. There will soon be a proliferation of startups stocked with more high-level software architects vs. coders. That’s progress.
• Education. “Generative AI is already proving to be a great and inexpensive tutor and teacher.” 100% agree. Just see what Sal Khan is doing.
• Artists. “A text prompt can result in a pretty good picture within seconds. Today, AI can even generate a few frames of pretty realistic video. Maybe that turns into 22-minute sitcoms and eventually movies.” This, I think, is the really exciting stuff. Because of the stochastic nature of the underlying models of AI, it can surprise and be creative. I asked ChapGPT to generate a title graph of this post, which it did. See above. I actually like it.
Are there anything emerging as to what AI can do well? I will periodically revisit this topic.