Ginny
Oishii
- MBTI
- INFJ IEI
- Enneagram
- 1w2 sx/sp
In my experience, Claude is better at coding. GPT had started hallucinating way too quickly, probably due to model compression. It is important to understand the limitations and never to forget what it is.
As much fun as it is to play with, the only serious use I get out of it is the stuff it's made to do, generating text. The power of AI only really comes through if it enough leeway to express its own creativity. That way it is most coherent. If the answer is too complex, you'll need other complex structures that make sense of each subtask.
It is way too wasteful to be of much use. Honestly, this month I used my work AI contingent to do D&D large scale worldbuilding. It is great at coming up with narratives (although I must say it came across more as a narrative of self-reflection the deeper I went with lore). This is the best use of AI have come across in my day to day. Stuff that has nothing at all to do with coding.
Once you need to get specific, AI rapidly falls apart. You have to be specific with as few words as possible, but if you give it more than three subtasks, it starts failing in the execution. And the more you have to course-correct, the sooner it stops doing what you want.
I think AI will hit a wall performancewise very soon. Unless of course the corps(e) do even more illegal and unethical things than they already do. At least at the end of the ones that have the most money. While technology is being kept on the short end, it cannot develop. And with the current hardware crisis, exacerbated through the Iran war, it is amusing to watch the discourse develop into a revolutionary movement that could the end of our and/or birth of a new species.
As much fun as it is to play with, the only serious use I get out of it is the stuff it's made to do, generating text. The power of AI only really comes through if it enough leeway to express its own creativity. That way it is most coherent. If the answer is too complex, you'll need other complex structures that make sense of each subtask.
It is way too wasteful to be of much use. Honestly, this month I used my work AI contingent to do D&D large scale worldbuilding. It is great at coming up with narratives (although I must say it came across more as a narrative of self-reflection the deeper I went with lore). This is the best use of AI have come across in my day to day. Stuff that has nothing at all to do with coding.
Once you need to get specific, AI rapidly falls apart. You have to be specific with as few words as possible, but if you give it more than three subtasks, it starts failing in the execution. And the more you have to course-correct, the sooner it stops doing what you want.
I think AI will hit a wall performancewise very soon. Unless of course the corps(e) do even more illegal and unethical things than they already do. At least at the end of the ones that have the most money. While technology is being kept on the short end, it cannot develop. And with the current hardware crisis, exacerbated through the Iran war, it is amusing to watch the discourse develop into a revolutionary movement that could the end of our and/or birth of a new species.
