A query last week suggested I might take part in an interview about artificial intelligence. I demurred. I do not consciously use AI, and I have reasons.
1. The material that AI uses has been plagiarized from many sources without respect for copyright.
2. Data centers that operate it are environmentally destructive and wasteful.
3. Reports abound of people who used AI for various purposes, only to wind up with material that was corrupt or fictional. It can't be trusted to deliver accurate information.
4. I am still waiting to see whether the whole hoopla is just another overextended bubble whose burst will turn the economy to tatters.
5. Major figures promoting AI are remarkably unlikable, and their motives and designs are suspect. I wonder whether they are similar to the private equity types who acquire properties, degrade the product, and extract more money from customers without giving value.
So I'm a mossback.
I am aware of colleagues who make use of AI to a limited degree, and others who are studying it to see whether it can be made useful and reliable, though I have not investigated the specifics.
I am in show-me mode. Show me that it actually works. Show me that it is fair. Show me that it is not harmful to the environment. Show me that it is reliable. Show me that it is not a scam. Show me that it is not harmful to people.
Anyone of show-me capacity who would care to instruct an elderly mossback are welcome to do so in the comments.
Insults and childish name-calling will, of course, be deleted.
Items 4 and 5 remain unclear to me. If we widen the concept of AI to encompass machine learning in general, then I doubt we're looking at a bust in the long term. Cars run on gasoline; AIs run on "compute." Right now, that's expensive and most of the companies providing them are doing so at a severe loss. So we either make users pay for compute, or we develop better models that run more efficiently.
ReplyDeleteNow, regarding items 1–3, my take is that the problems you identify stem from users who don't understand how to use AI. For example, I needed a spreadsheet with rows and columns labeled varies things and some cells needed complex formulas. I could have done this myself, but it would take an hour or two. An AI knocked it out in less than a second. Perfect, flawless. Because that was a mechanical task. AIs are good at that. It's when people try to use it to make value judgements, render opinions, areas of subjective analysis that AIs start to show their limitations. "Write an enjoyable children's story" will generate slop. "Read my children's story and recommend replacements for any 4- or 5-syllable words" is more in line with what the AI can do, and keep the user in control of the output.
I'm not an AI evangelist. But I work in publishing and have felt the discomfort (terror) of this technology and understand that we can't wish it away. Certain tasks, such as TOC and index construction are in the human realm, yet an AI could probably yank out enough material to give you a running start. But I would still do it manually. However, "Find me all the publishing agents who accept this kind of material. List them in order of prominence. State their employers, and give me their email and mailing addresses, and their phone numbers."
That would return good results and save time, and be useful if you were shopping a manuscript.
It doesn't work, except perhaps for nefarious purposes like brute-force hacking and financial scams. It is not "intelligence".
ReplyDeleteI used Claude a few times, because it was free.
I sought help on a rather complicated text-based game I have worked on for fun, one with many functions, options and a display. I asked it to make the display better. Claude made it worse.
I have a series of mystery stories that I am revising from 3rd person to 1st person. I submitted one and asked Claude to do the revision. It basically didn't get it. The revision was lifeless, the characters did not act properly, the solution was never really stated. No drama and no fun.
I asked about a science fiction story that I was writing. It seemed to be interested in the concept, but it only came up with things I had already imagined. When I revealed the end to it, Claude acted very displeased and wanted it changed. The aliens want our youngest and brightest as trainees. One of the trainees writes home before visiting, and discloses that he and another trainee (a female) are now partners and sharing their quarters on the alien space ship. Claude was almost prudish about that, because they were 16 and 15 years old. This was amusing, but it was the last straw.