AI-yi-yi

[Flex]

Here's a really long but good thread on DeepSeek and what this news actually means. There are a lot of potentially major environmental implications, among other things, so probably worth taking the time to at least understand a little bit even if you're just genuinely AI skeptical (as I am): https://bsky.app/profile/karenhao.bsky. ... q4pmrwgl2v

[matedog]

Also Bonham died in 1980 and the banned formed in 1981.

Using Bing's free ai image generator (they don't allow for generating real people) :

[Dr. Medulla]

What, a chatbot might have built-in restrictions or set responses based on political interests? Say it ain't so, Joe, say it ain't so!
https://www.theguardian.com/technology/ ... and-taiwan

[Dr. Medulla]

Marina Hyde enjoys the tech bros' tantrum about DeepSeek: https://www.theguardian.com/commentisfr ... man-openai

[Silent Majority]

https://arstechnica.com/tech-policy/202 ... ntent=null

Trapping AI scrapers.

[Dr. Medulla]

https://arstechnica.com/tech-policy/202 ... ntent=null

Trapping AI scrapers.

Fascinating and depressing all at the same time. Yay for malware!

[gkbill]

https://arstechnica.com/tech-policy/202 ... ntent=null

Trapping AI scrapers.

Fascinating and depressing all at the same time. Yay for malware!

Hello,

If it leads to crawlers having to pay for crawling, it's good.

[Dr. Medulla]

https://arstechnica.com/tech-policy/202 ... ntent=null

Trapping AI scrapers.

Fascinating and depressing all at the same time. Yay for malware!

Hello,

If it leads to crawlers having to pay for crawling, it's good.

My concern is more the proliferation of malware. Even if applied toward other malware, that just ups the amount of destructive stuff out there.

[Dr. Medulla]

https://theonion.com/nbc-producers-deny ... eles-show/

[Silent Majority]

Use it or lose it. Same as it ever was.

https://www.404media.co/microsoft-study ... repared-3/

[Dr. Medulla]

Use it or lose it. Same as it ever was.

https://www.404media.co/microsoft-study ... repared-3/

But thinking is hard!

[Silent Majority]

https://financialpost.com/technology/te ... ta-centres

Not a bubble popping, but perhaps the build up to that.

[Flex]

So, I'll admit that I recently used ChatGPT for a couple work items and the missus has used it as well for a job she's applying for. Our experiences, briefly:

-My wife is in the final stages for a new job, external from her company. It would be a promotion, which is great. She hadn't updated or worked on her resume since, well, her last job so she thought she'd give ChatGPT a try to basically give it a refresh and help align her resume, cover letter, etc. to the job she was applying for. She was able to feed the AI all of her old material, the job description, and a bunch of metrics and shit that she has from her current role and got ChatGPT to give her a very good draft of a resume and cover letter. She still had to work on it further herself, and it required feeding a LOT of info upfront to get a good draft back, but it probably saved her hours of work. So that was an interesting use case. She's continued to use it during the application process, brainstorming interview strategies and questions, stuff like that. That ALL still requires a lot of initial data input, but I can see it's value.

-I had two work use cases where I used ChatGPT. One was creating a simple comms plan for a client. Extremely small scope, just a quick draft for them to do with what they will. I spent some time providing parameters to ChatGPT - a zero draft/outline, essentially - and from that ChatGPT worked that up into a solid high-level plan which was all I was looking for. I still had to have a passthrough of it to adjust things to my liking and input some more personal knowledge/insight, but having it basically build out a plan framework and get it to a first draft probably saved me a couple hours of work. Since I wasn't be paid hourly on this: totally worth it.

-One other use I tried was doing some research for some names of folks with subject matter expertise we wanted to invite to an event for a client. Again, not a huge project for me, but I thought I'd give the research component of ChatGPT a try. Again, I spent some time carefully thinking through and setting parameters of what I was looking for and ChatGPT served me up a solid list of names in the number range I was looking for. I then went through and googled each proposed name. Almost all of them were correct in terms of background/expertise and that they'd be of interest to the client. Some I had to fill out contact information that seemed readily available but that for some reason ChatGPT said wasn't, and a few government roles had outdated people in their positions that I had to research myself and update. But it was pretty good, I definitely got some sets of names that were good value adds for the client that I wouldn't have gotten with random googling.

So, I tried another round to see if I could net out a few more names and this time every single name I got back was fake. Fake names, fake titles, fake organizations, all of it. That's sort of what I was expecting more of from the first go-round, but it's like you got one bite at the apple for good info and if you pressed, ChatGPT would just make shit up.

I'll be honest: I can see it's value as an initial-draft-generator or as a light researcher. That's not nothing, it saved me some hours of work and I got some results I probably wouldn't have otherwise, but it's also hardly revolutionary. And the key is that it still required me to have a good understanding of what I wanted to get out of it and have the expertise to review what it spit out. And my wife had a lot of data to feed in and very clear ideas of what she wanted to get out of the program. That's useful, but it doesn't feel revolutionary. And ChatGPT obviously comes with a lot of negatives that make it a tough value-proposition to be able to use ethically. But I figured I had to see what the fuss was about.

[Dr. Medulla]

So, I tried another round to see if I could net out a few more names and this time every single name I got back was fake. Fake names, fake titles, fake organizations, all of it. That's sort of what I was expecting more of from the first go-round, but it's like you got one bite at the apple for good info and if you pressed, ChatGPT would just make shit up.

I read stories of law profs playing around with it by having it summarize relevant cases for particular issues and found it generated amazing summaries of cases that flat out didn't exist. It's like disinformation is built into its DNA.

[Flex]

So, I tried another round to see if I could net out a few more names and this time every single name I got back was fake. Fake names, fake titles, fake organizations, all of it. That's sort of what I was expecting more of from the first go-round, but it's like you got one bite at the apple for good info and if you pressed, ChatGPT would just make shit up.

I read stories of law profs playing around with it by having it summarize relevant cases for particular issues and found it generated amazing summaries of cases that flat out didn't exist. It's like disinformation is built into its DNA.

I assume chatgpt will be getting a high level cabinet position in the trump administration shortly

Faux addendum: musk is almost certainly using AI on all of the mass layoff/terminations.