HawkinsHollow
Veteran Member
CBS news just out this spot on AI a couple of days ago.
AI and it's usefulness in everyday life?
- Thread starter paulsharvey
- Start date
- Views: 3132
ericm979
Super Member
- Joined
- Nov 25, 2016
- Messages
- 5,743
- Location
- Santa Cruz Mountains CA, Southern OR
- Tractor
- Branson 3725H Deere 5105
I avoid it as much as possible. I'm not being a Luddite; my career was in tech. I have a pretty good understanding of how current "AI" works.
Current AI LLMs have a lot of problems:
-they're trained on material illegally scraped from the internet in violation of copyright law
-they're set up to emit plausible sounding information to a prompt. That information being accurate is not a goal nor is it possible with current LLMs
-they're being used to produce garbage writing to replace stuff (i.e documentation) that used to be written by humans who at least tried to be accurate.
-they're massively over hyped. They're not AGI or even close.
-bosses and investors are excited because they hope to replace employees. What is that going to do to society?
-current LLMs are not AI- they're not intelligent. They have no thought or understanding of meaning. Its just statistically generated tokens.
The kind of pattern matching that's used in LLMs does have some real uses. For example identifying cancer cells in imaging, or identifying small wildfire starts on camera. But it's not intelligence.
Current AI LLMs have a lot of problems:
-they're trained on material illegally scraped from the internet in violation of copyright law
-they're set up to emit plausible sounding information to a prompt. That information being accurate is not a goal nor is it possible with current LLMs
-they're being used to produce garbage writing to replace stuff (i.e documentation) that used to be written by humans who at least tried to be accurate.
-they're massively over hyped. They're not AGI or even close.
-bosses and investors are excited because they hope to replace employees. What is that going to do to society?
-current LLMs are not AI- they're not intelligent. They have no thought or understanding of meaning. Its just statistically generated tokens.
The kind of pattern matching that's used in LLMs does have some real uses. For example identifying cancer cells in imaging, or identifying small wildfire starts on camera. But it's not intelligence.
EddieWalker
Epic Contributor
I haven't used it unless we count auto correct, which seems to be a lot better then it used to be. Especially with punctuation.
My wife teaches at the local university. At first, students were using it to cheat, but then the teachers figured out how to use AI to detect cheating. It's very good at that, especially plagerism. It rates every paper on what percentage of it has been used before in other papers. Most are about ten percent, but the cheaters always get carried away and they come in at 80 to 100% copied from other papers.
What she likes the most about it is in writing questions for exams. There is a massive black market for exam questions and answers, so they have to write new questions for every test. AI has made this a lot easier.
My wife teaches at the local university. At first, students were using it to cheat, but then the teachers figured out how to use AI to detect cheating. It's very good at that, especially plagerism. It rates every paper on what percentage of it has been used before in other papers. Most are about ten percent, but the cheaters always get carried away and they come in at 80 to 100% copied from other papers.
What she likes the most about it is in writing questions for exams. There is a massive black market for exam questions and answers, so they have to write new questions for every test. AI has made this a lot easier.
Oaktree
Super Member
I wouldn't say that, sure makes it easy when you want to know how to do something to be able to Google it and have the answer in less than a second. I do think the "internet" has contributed to the polarization in the country today...there's less and less need to associate with people you don't agree with. I don't see that as a good thing.
I'll be the first to admit I really don't understand it, nor do I have the slightest idea how you'd use it (at least knowingly...I'm sure something like a Google search involves AI in some form)
"Cherish" wasn't a 70s song, it came out in the summer of 1966.
Today in 1966: The Association’s “Cherish” Hits the Hot 100 | Rhino
52 years ago today, The Association’s most durable hit single made its debut on the Hot 100. That’s not to say that it’s your personal favorite, but given how successful it was at the time of its release and how many times it’s been covered over the years, not to mention how much airplay it...
www.rhino.com
CalG
Super Member
- Joined
- Sep 29, 2011
- Messages
- 5,458
- Location
- vermont
- Tractor
- Hurlimann 435, Fordson E27n, Bolens HT-23, Kubota B7200, Kubota B2601
The BIG issue that concerns me is how willing people (sheep) are to accept whatever ai spits out as the god's honest truth and the "only way".
And yet it can only spit back what goes in.
Wikipedia is like that, but,,,,
Look! A squirrel!
And yet it can only spit back what goes in.
Wikipedia is like that, but,,,,
Look! A squirrel!
otlski
Gold Member
I have interacted with AI in a few ways. Before getting into that, let me say that at this point I would not trust AI to opine or report on any controversial topic. That said, doing such was my first real interaction.
At the beginning of the Ukraine Russia war I chatted with an AI asking its opinion on what was happening. I choose that topic specifically because people had a hard enough time discussing it, so the mark of a good AI might be to pull off the conversation well. It was not. The chat seemed very forced. The AI was full of "facts" and insisted in hedging every word. Seriously noncommittal. I tried to force it down more committed avenues, but in the end it simply regurgitated and added nothing of value to the "conversation".
In another case, I wanted to test AI's ability to write marketing content on a highly technical subject. I gave it the topic, keywords to include, and the article length. It did a decent job. It still read a little "clickbaity" but was better than my worst writers but not as good as the best. Overall, it gave a good head start so that a good writer would save hours on the initial draft and just had to wordsmith the rest of the way through. By way of comparison, a good writer would prefer to start fresh than to have to wordsmith initial copy from a bad writer.
In another case, I asked a niche technical question that I already knew the answer to. The answer was astoundingly good. Consise, highly technical; correct in every meaningful way. Then I looked at the links it used to build its consensus answer and the links in four out of five cases were to my company's website's white papers. Hmmm. That's why the answer was spot on. Good I suppose... but I am not quite sure.
Finally, my CEO was beating a dead horse longstanding argument we've had. In it, he made a comparison between my business unit to one of our 90 sister companies. During the debate the CEO said "but blankety blank sells blanks all the time and they know blank-blank ahead of time, you are telling me you don't?". So I typed up his statement into the AI as "How many blanks does blankety blank sell per year?" just to get a scope for comparison. The AI replied "blankety blank does not sell blanks per se, instead they sell management services for these things. It went on to say the first one-off product in the 1950's was indeed a blank but that transition into admin and logistics. So in the end it gave me the arrow in the quiver to hold the CEO at bay (I did not win).
Draw your conclusion on this as you will.
At the beginning of the Ukraine Russia war I chatted with an AI asking its opinion on what was happening. I choose that topic specifically because people had a hard enough time discussing it, so the mark of a good AI might be to pull off the conversation well. It was not. The chat seemed very forced. The AI was full of "facts" and insisted in hedging every word. Seriously noncommittal. I tried to force it down more committed avenues, but in the end it simply regurgitated and added nothing of value to the "conversation".
In another case, I wanted to test AI's ability to write marketing content on a highly technical subject. I gave it the topic, keywords to include, and the article length. It did a decent job. It still read a little "clickbaity" but was better than my worst writers but not as good as the best. Overall, it gave a good head start so that a good writer would save hours on the initial draft and just had to wordsmith the rest of the way through. By way of comparison, a good writer would prefer to start fresh than to have to wordsmith initial copy from a bad writer.
In another case, I asked a niche technical question that I already knew the answer to. The answer was astoundingly good. Consise, highly technical; correct in every meaningful way. Then I looked at the links it used to build its consensus answer and the links in four out of five cases were to my company's website's white papers. Hmmm. That's why the answer was spot on. Good I suppose... but I am not quite sure.
Finally, my CEO was beating a dead horse longstanding argument we've had. In it, he made a comparison between my business unit to one of our 90 sister companies. During the debate the CEO said "but blankety blank sells blanks all the time and they know blank-blank ahead of time, you are telling me you don't?". So I typed up his statement into the AI as "How many blanks does blankety blank sell per year?" just to get a scope for comparison. The AI replied "blankety blank does not sell blanks per se, instead they sell management services for these things. It went on to say the first one-off product in the 1950's was indeed a blank but that transition into admin and logistics. So in the end it gave me the arrow in the quiver to hold the CEO at bay (I did not win).
Draw your conclusion on this as you will.
bigtiller
Super Member
- Joined
- Feb 1, 2006
- Messages
- 6,343
- Location
- central Iowa
- Tractor
- John Deere 2720 John Deere 3039R John Deere Z545R
Yep, that is exactly what the A.I. said to me.
For my crowd of friends, it was a very popular song well into the 70's.
CalG
Super Member
- Joined
- Sep 29, 2011
- Messages
- 5,458
- Location
- vermont
- Tractor
- Hurlimann 435, Fordson E27n, Bolens HT-23, Kubota B7200, Kubota B2601
Recent commentary
Just like the internet, a mile wide and an inch deep.
I'm curious, how many of us feel that a "search engine", such as google, Duck Duck go, or Brave, are the same as "artificial intelligence"?
Just like the internet, a mile wide and an inch deep.
I'm curious, how many of us feel that a "search engine", such as google, Duck Duck go, or Brave, are the same as "artificial intelligence"?
OP
Hmm, yes and no. Sure, the search engines are using it to a small degree, but its not the same. Deepseek, you can get 10 recipes for tonight, to feed a family of 5, using 3 chicken breast's, and it will give you 4 or 5 whole meals, with the recipe, cooking instructions, and a "why it works". Then, at the end of that, you can type, "We don't like cumin, or I'm allergic to peanuts", it will refine the recipes suggested.
You don't type it in Google and get list of recipes. You get a list of websites that might have recipes, and likely not what you asked for, then you search around on the web pages, to find they aren't what you wanted.
So, yesterday, not what I would call productive, more playing; i decided to try it's ability to narrow down suggestions. "What rubber worm color should work in Lake Santa Fe on large mouth bass in late spring and early summer". It suggested June bug, green watermelon, and 2 or 3 others, as well as sizes, and the Why; water is clearer than average, and those more natural colors are more visible to fish then in many of the dirtier lakes.
Yes, Lake Santa Fe is a pretty good sized lake, but Google/Yandex/Whatever isn't going to give you nearly as localized advice/or the reasons.
For the record, I do find June-bug to be the most effective color...
Last edited:
OP
As mentioned by someone, I dont think you want to get all philosophical with it, or ask it of rain turtles work. Ask it about rain turtles, and it's likely to suggest other folk magic to get it to rain, like nailing a snake skeleton to a tree...