ChatGPT and AI is at the forefront of the US investing community at the moment. As an investor who has a background in IT with some AI knowledge and has studied some simple Machine Learning aspects at university such as Neural Networks and Genetic Algorithms, I thought I’d weigh in with some thoughts on the subject.
Because of all the AI hype nascent from ChatGPT, there’s a chance that ChatGPT could break the internet… Not in like a SkyNet AI way nor in a slang way nor due to excess traffic, but because it could cause people to stop sharing information.
Some years ago now, Google improved their search engine by collecting data from websites and displaying answers to people’s questions without having to click on links and go to websites. For example, if you search for “what is planck’s constant?” Google will give you the answer straight away. No need to trawl through websites for the answer.
They did limit this functionality quite significantly though for fear of breaking the internet. Because this information comes from websites, if Google makes it so you don’t need to go to those websites that produce or share the information, those websites will get no traffic. If those websites get no traffic, they won’t be able to sell products or collect advert revenue. This will cause them to shut down, which in turn will cause Google to have less information feeding it and eventually break the internet – no new information being uploaded (at least not without a paywall – which would break search engine algorithms because you wouldn’t know which paywall had the info you needed).
ChatGPT runs a similar risk as it’s trained by data from websites to give people search result information without the user needing to go to the website hosting the data. My concern is that the likes of Microsoft (who are desperate to compete against Google in the search and advert revenue arena) will make a successful AI search engine that does exactly that and breaks the internet.
I’ve used ChatGPT and was very impressed with it. However, talking to others who’ve tested it for other things, suggests that it’s probably not very good for all types of information and it fabricates a lot of information. So in conclusion, I think the internet is safe for now, but we do run the risk of someone significantly changing the culture and nature of the internet in the future.
As for ChatGPT and the AI trend, I think it definitely has it’s uses. Particularly in places where the impact of getting the job wrong is low, the tasks are simple enough to do but require a highly capable communicator or need a gimmick. Taking orders at the drive through, with the option of a human override might be such a task. Outside of this, I can imagine more complex uses for AI, such as coding where the parameters within which the coding may be done are very well defined and possibly constrained (or not, if there’s enough work put into the project).
In summary, a lot of thought needs to be put into the application of AI, specifically around it’s cost and suitability for any given solution.