My favorite part about this (and all GenAI) comments section is where one person says, "This is my personal experience using AI" and then a chorus of people chime in "Well, you're using it wrong!"
I personally prefer the one where everyone tells you that your error is because you used the outdated and almost unusable version from yesterday instead of the revolutionary release from today that will change everything we know. Rinse and repeat tomorrow.
Not to mention the variations of "you need to prompt better" including now "rules files" which begs the question: wouldn't just writing code be a much better way to exercise control over the machine?
In my tests of using AI to write most of my code, just writing the code yourself(with Copilot) and doing manual rounds with Claude is much faster and easier to maintain.
One very irritating problem I am seeing in a bunch of companies that I have invested in and my money is at stake is that they have taken up larger investments from normal VCs who are usually dumb as rocks but have a larger share is that they are pushing heavily for AI in the day to day processes of the company.
For example, some companies are using AI to create tickets or to collate feedback from users.
I can clearly see that this is making them think far less through the problem and a lot of this sixth sense understanding of the problem space happens through working through these ticket creation or product creation documents which are now being done by AI.
That is causing the quality of the work to become this weird drone like NPC like state where they aren't really solving real issues yet they're getting a lot of stuff done.
It's still very early so I do not know how best to talk to them about it. But it's very clear that any sort of creative work, problem solving, etc has huge negative implications when AI is used even a little bit.
I have also started to think that a great angel investment question is to ask companies if they are a non AI zone and investing in them will bring better returns in the future.
It’s because it’s never “this is my personal experience”, it’s always of the “this whole AI thing is nonsense because it doesn’t work for me” variety.
The same can be said about the other side. It is rarely phrased has "LLM is a useful tool with some important limitations" but "Look the LLM manage to create a junior-level feature, therefor we won't need developers in 2 years from now".
It tends to be the same with anything hyped / divisive. Human tend to exaggerate in both direction in communication, especially in low-stake environment such as an internet forum, or when they stand to gain something from the hype.
You seem to have confused "The whole AI thing is nonsense. [Anecdote]." with "The whole AI thing is nonsense because [anecdote]." I see a lot of "LLMs are not useful. e.g. the other day I asked it to do X and it was terrible." That is not somebody saying that that one experience definitively proves that LLMs are useless, or saying that you should believe that LLMs are useful based only on that one anecdote. It is people making their posts more interesting than just giving their opinions along with arguments for those opinions.
Obviously their views are based on the sum of all their experience with LLMs. We don't have to say so every time.
It's because everyone's "personal experience" is "I used it once and it didn't work".