Post AV8UXLwFRrfWLIlcae by titinoNero@mastodon.uno
(DIR) More posts by titinoNero@mastodon.uno
(DIR) Post #AV8NNFfDDvowJ52EMK by quinta@mastodon.uno
2023-04-29T05:17:26Z
0 likes, 1 repeats
ChatGPT legal disclaimer:"Services like ChatGPT generate responses by reading a user’s request and then predicting the next most likely words that might appear in response. In some cases, the most likely next words may not be the most factually accurate ones. For this reason, you should not rely on the factual accuracy of output from our models"what should we rely it for, then ?
(DIR) Post #AV8NjLsimoelxhtMcS by elCelio@mastodon.uno
2023-04-29T05:21:28Z
0 likes, 0 repeats
@quinta so it behaves like a human being 😲
(DIR) Post #AV8UXLwFRrfWLIlcae by titinoNero@mastodon.uno
2023-04-29T06:37:44Z
0 likes, 0 repeats
@quinta rely it as "spercazzola" generator. As "fuffa" fillers.
(DIR) Post #AV8Uub0JbK8fjCkiUS by Davide_Sandini@mastodon.uno
2023-04-29T06:41:56Z
0 likes, 0 repeats
@quintaQuindi come suggeriva qualcuno scrivere "stronzate artificiali" vicino ad ogni occorrenza di "intelligenza artificiale" le fará come per magia diventare sinonimi anche nei loro output?
(DIR) Post #AV8UyHGquWqBgWF8vA by edo78@mastodon.uno
2023-04-29T06:42:35Z
0 likes, 0 repeats
@quinta there are a shitload of use cases in which #chatgpt is useful even if it's not factually accurate.For example it's good at generating text and it's not so bad at explainning things. Like wikipedia it's not to be relied on without checking but it's useful nontheless
(DIR) Post #AV8VZnJQkAyHm5L2rw by Velveteen@snowdin.town
2023-04-29T05:34:53.165370Z
0 likes, 1 repeats
@quinta you shouldn't rely on it for anything for any reason basically. Treat it like a toy and take nothing at face value
(DIR) Post #AV8VaupOHqXFNBp2lU by monetsonante@mastodon.uno
2023-04-29T05:35:12Z
0 likes, 1 repeats
@elCelio @quinta Forget the intellegent part (it's not). I think is a really useful tool, just like a excel on steroids. A way to generate text (and texts are often ideas) and mix, convert, translate, refine them.
(DIR) Post #AV8Vtc6q6e9MovtqN6 by quinta@mastodon.uno
2023-04-29T06:52:52Z
0 likes, 0 repeats
@edo78 I don't agree. it's orders of magnitude less reliable than Wikipedia, which is thoroughly checked (which doesn't exclude errors, bias, etc... but it's checked)it's good at generating theft which is not trustworthy. (aka likely-bs)
(DIR) Post #AV8VzqkYaNH5xtS0rw by quinta@mastodon.uno
2023-04-29T06:54:07Z
0 likes, 0 repeats
@edo78 please provide 10-20 examples of that shitload of use cases. I can imagine just few.
(DIR) Post #AV8Xxh5fjCcTKtQXXU by edo78@mastodon.uno
2023-04-29T07:16:02Z
0 likes, 0 repeats
@quinta sure thing.I use it to generate some ideas (and have some feedbacks on mine) about IT solutiones from shell scripts to home automation.I have it write some short novel tailored to my liking. I have it providing feedbacks on my own writing, from novels to tech documentations.I have it provide an explaination of complex systems like the JSON exported from nodered and I have it modified to match my goals.I come back to old discussion asking for improvements or clarification and so on...
(DIR) Post #AV8YBCETFpzLgYysxU by edo78@mastodon.uno
2023-04-29T07:18:32Z
0 likes, 0 repeats
@quinta sure, I doesn't want to imply that chatgpt is as reliable as wikipedia just that you shouldn't rely only on one source.To be honest I pretty much hate how confident chatgpt sounds even after you point out to it the Nth error it committed
(DIR) Post #AV8gfOsXbEePECdXlI by quinta@mastodon.uno
2023-04-29T08:53:39Z
0 likes, 0 repeats
@edo78 novel is fiction, no problem there. generate ideas is ok; for the rest, I wouldn't rely on it
(DIR) Post #AV8n6RvUyYgaDxIkme by edo78@mastodon.uno
2023-04-29T10:05:47Z
0 likes, 0 repeats
@quinta I don't trust it 100% as I don't trust stackoverflow but it's a freaking good starting point
(DIR) Post #AV91TddD1P12XrhLSC by cage@stereophonic.space
2023-04-29T11:58:12.859611Z
0 likes, 0 repeats
@quinta>> ChatGPT legal disclaimer:[...]> what should we rely it for, then ?I tried to address exactly this kind of (easy, admittedly) criticism,"ChatGPT makes too many mistakes to be useful", to a friend of mineand their answer surprised me, someway. What they said to mesummarized to: "ChatGPT is going to be useful as long as it will allowentrepreneurs to save money by lowering the number of their employers(or their pay); as long as the saving is good enough, quality of itsoutput does not matter".So far i was not able to elaborate anything as rebuttal to thisargument, under the assumption of a capitalistic economic system.Bye!C.
(DIR) Post #AV91TeRC1Yzx2t5GMa by amiloradovsky@stereophonic.space
2023-04-29T12:46:50.197558Z
0 likes, 0 repeats
@cage @quinta lowering the quality for the price and quantity means leaving one market segment and entering another, it's changing the business model, not something an entrepreneur may want to dothe benefit of such a systems may be assisting the humans in their trade, not art, where volumes of unoriginal standard text (boilerplate) need to be generated first, then edited manuallyat the end of the day, it's still a human coming up with the request/prompt and rectifying the response, and then taking the responsibility for the end resultjust like machine-generated translation, for example
(DIR) Post #AV91qpa4lAwaGZI9DM by quinta@mastodon.uno
2023-04-29T12:50:45Z
0 likes, 0 repeats
@cage liability? consumer protections? shareholder protection rules?
(DIR) Post #AV97Oubn4b0jTZAF9M by cage@stereophonic.space
2023-04-29T13:32:04.545464Z
0 likes, 0 repeats
@amiloradovskyHi!> @cage @quinta lowering the quality for the price and> quantity means leaving one market segment and entering another, it's> changing the business model, not something an entrepreneur may want> to doYes but, in general, I think this is true just after some kind of"quality lower threshold", so to say, has been trespassed (maybe somehard limits like government's market regulation¹, or just quality isbecome so low that is harmful for the company's profits).> the benefit of such a systems may be assisting the humans in their> trade, not art, where volumes of unoriginal standard text> (boilerplate) need to be generated first, then edited manually at> the end of the day, it's still a human coming up with the> request/prompt and rectifying the response, and then taking the> responsibility for the end result just like machine-generated> translation, for exampleI think I can see the point and I mostly agree, but now my questionsare: how many people are needed for this work of AI output refinementto be done (let's say -just kidding- generating marketing fediverseposts 😃) comparing to the non AI assisted job? And even if the samenumber of persons are needed, would not the pay for each employer belowered, at least for number of posts produced (AI supposed to makepossible to do the same job in less time)?Bye!C.¹ not taking account lobbying, conflict of interest or, even worse,corruption
(DIR) Post #AV97OvD0qCmjL0ZyAS by amiloradovsky@stereophonic.space
2023-04-29T13:53:13.312507Z
0 likes, 0 repeats
@cage @quinta re market segments: no, it's not just government regulation; I for example have a quality threshold, depending on my budget, if it's cheaper for the same quality then great, then I could either afford higher quality or save some money, but I wouldn't just buy something because it's cheap(er than before)re "AI" assistance: the general rule is that due to technology the number of jobs is fewer but qualification requirements are higher (therefore the pay)
(DIR) Post #AV9Ou0AIA9EiuHiEXQ by cage@stereophonic.space
2023-04-29T16:57:30.484193Z
0 likes, 0 repeats
@amiloradovskyHi!> @cage @quinta re market segments: no, it's not just> government regulation; I for example have a quality threshold,> depending on my budget, if it's cheaper for the same quality then> great, then I could either afford higher quality or save some money,> but I wouldn't just buy something because it's cheap(er than before)I agree with you here, i tried to write the same idea in my parentmessage but i missed a few word one o my phrases, sorry.> re "AI" assistance: the general rule is that due to technology the> number of jobs is fewer but qualification requirements are higher> (therefore the pay)OK, only I do not know if this rule apply in the case of chatGPTadoption, I am a bit pessimistic because in the country where I live,despite the technology innovation (and i suspect the productivity),the employer's pay are decreasing. Anyway we'll see and I hope I amwrong! :)Bye!C.
(DIR) Post #AV9Ou0q7echH01Hdjs by amiloradovsky@stereophonic.space
2023-04-29T17:09:19.552423Z
0 likes, 0 repeats
@cage @quinta to me ChatGPT and GPT-N's are completely uninteresting, just another Elisa, IMHOspeaking of the decreasing pay, I guess it depends on the occupation: somewhere it's up and somewhere it's down, due to the technological advancesone has to either change the occupation or invest into new tools (plus the knowledge and skills of using them effectively) -- after tractors have been invented, having and using them became indispensable for farmers, but farming itself didn't go away, the rest of the people formerly employed in agriculture became industrial workers -- this is the same thing today, but industrial workers turn into programmers
(DIR) Post #AV9Ph6QVfetuW4iAQi by amiloradovsky@stereophonic.space
2023-04-29T17:18:12.605443Z
0 likes, 0 repeats
@cage @quinta *ELIZA
(DIR) Post #AV9RIvCJy3wkeIfj3w by naufraghi@mastodon.uno
2023-04-29T17:36:12Z
0 likes, 0 repeats
@quinta I used ChatGPT to create stories of elves, reindeer, and Santa Claus for my children in December. The model is pretty good at this and gave me some good ideas (probably taken from a story it found online and I will never know who to thank).
(DIR) Post #AV9YWUUWl1TpYPbaT2 by cage@stereophonic.space
2023-04-29T18:49:21.994761Z
0 likes, 0 repeats
@amiloradovskyHi!> to me ChatGPT and GPT-N's are completely uninteresting, just another Elisa, IMHOThis appears to be a bit unfair to me, also i think a lot ofinteresting philosophical questions could arise from yoursentence. But I am afraid this is not the place for this kind ofdiscussion.To me chatGPT write a better English than mine ad has a lot of -mostlywrong- opinions about everything, like many people I know! 😃😆> speaking of the decreasing pay, I guess it depends on the> occupation: somewhere it's up and somewhere it's down, due to the> technological advances one has to either change the occupation or> invest into new tools (plus the knowledge and skills of using them> effectively) -- after tractors have been invented, having and using> them became indispensable for farmers, but farming itself didn't go> away, the rest of the people formerly employed in agriculture became> industrial workers -- this is the same thing today,Again, I am not sure this time will be the same, but I have no strongopinion, i just wait and see what will happens.> but industrial workers turn into programmersSo true!Bye!C.
(DIR) Post #AV9YWV3GfrGlI9rKcK by amiloradovsky@stereophonic.space
2023-04-29T18:57:06.886526Z
0 likes, 0 repeats
@cage OK, I only want to add that it's important to emphasize that a chat(ter)bot doesn't have opinions, it merely repeats after the train data, that's all it does
(DIR) Post #AV9YdXtSVU548HdAUy by quinta@mastodon.uno
2023-04-29T18:58:21Z
0 likes, 0 repeats
@naufraghi lots of copyright issues here..
(DIR) Post #AV9jkazGVBnPFZuXD6 by cage@stereophonic.space
2023-04-29T19:52:15.045793Z
0 likes, 0 repeats
@amiloradovsky> OK, I only want to add that it's important to emphasize that a> chat(ter)bot doesn't have opinions, it merely repeats after the> train data, that's all it doesSure, and the latest systems also needs a lot of training data and, Iguess, a lot of energy is spent in both training and generatingprocess too.Bye!C.
(DIR) Post #AV9jkbhDrkxRRuTdj6 by amiloradovsky@stereophonic.space
2023-04-29T21:02:55.903813Z
0 likes, 0 repeats
@cage yep, training is very expensive for something like the GPTs, but inference is way cheaper