For dismantling the education system, as far as I can tell. The most common use case I hear for it is young people using it to write essays and homework, probably pretty soon theses, dissertations, and scientific studies. I'm already having to ban my tech savvy 11yo from it.
I don't really buy the idea that people are going to use ChatGPT to write theses, dissertations, and scientific studies. Maybe they will use them to polish the English prose, but if the implication is that they're hallucinating the work: no, that won't happen at large.
there's always been fraud in science--ChatGPT will further accelerate a particular kind of fraud, for sure. But I don't think we're anywhere close to that happening en-masse, tbh--if the work is not actually correct, and did not follow serious research methods--that's still a serious flaw.
I have observed that native English speaking students tend to paraphrase/edit ChatGTP output in such a way that it is difficult to be sure the work originated with ChatGPT.
Whereas, non-native speaking students tend to submit unedited ChatGPT output and thus are more easily detected.
During my last few semesters in university, I had teachers making sure that ChatGPT would answer wrong to their exam and homework questions (because ChatGPT is really bad at anything science-related, so that's not too hard to do), but programs relying on dissertations were in crisis mode.
Have no answer to that. But to be fair, it had some use for me (a lawyer, and not IT guy) to create code in Python when I wanted to work on some data for my private curiosity (like counting mandates in d'Hondt system or analysing temperatures during last decades in my local meteo station).
It's the 2024 equivalent of Clippy but much slower and I have a lot evangelists telling me it will replace everything. Yet the people saying that have to carefully orchestrate demos to just get it to do things.
I use it for summarizing 4 hour long meeting blocks to give me the gist of what happened. It does an OK job. Asking it for a succinct list of action items though? Terrible. It can be a time saver but you already have to know what you are expecting going in and adjust for it.
Hey man I just used it to write C++ code to compute the area of a polygon. Probably saved me 5 minutes. That's billions of dollars (of other people's money obv) well spent.
If you are a college graduate with a comp sci degree and a few small coding projects under your belt, an LLM chat bot can bring the same value to a company as you and everything you learned in school.
my PhD students, who write top papers and do seriously hard work, regularly use ChatGPT to accelerate / complement their work. There are serious issues with ChatGPT (it's all text, no semantic guarantees *at all*), but anyone saying it has zero applications has their head in the sand.
How else am I supposed to role play with Sean Connery as John Patrick Mason from The Rock (1996)? How else can I ask Patrick Star whether he thinks Picard or Kirk is the better Enterprise Captain? These are the hard hitting questions people demand answers for.
openAI's transformer models are really impressive from a technical standpoint, but i do think we would be better off if chatGPT never advanced past a gpt-3 api demo and didn't get very popular.. as great as llms can be, they've probably done more long-term worse than good in my opinion
For me it's replaced the Q&A aspect of search engines. Google was meant to search for documents on the web, not to answer questions. ChatGPT/Copilot/Gemini that have access to the web do a way better job at giving a direct answer to my questions.
What’s even more fun is that those who said back then that it makes them 10x developers still haven’t created anything that’s not a GPT wrapper in the equivalent of 20 dev years or more (the AI didn’t just stagnate).
But that guy with the goat face told me that it was important and would transform the world! And then when I went to make fun of it on Reddit, all the dudes with diamond-hands avatars said that I was a knuckle dragging Luddite. So surely, SURELY it must be very important and worthwhile!
I just got laid off from a nonprofit where a bunch of VC-brain dorks on the board are convinced we're going to replace every social worker with AI in like 3 months. It's embarrassing.
In 3 months, no. In the long term though? They’ll probably get their way, either by drastically reducing or outright eliminating an entire sector of jobs. No, AI won’t do the job as well as a human, but that’s not the point
Oh I get it, they're trying to do that now. We'd already been told not to "waste" our time writing emails and contacts to clients but just letting AI write them. You know, just what people who are struggling with their mental health need.
That sucks ass, bud. I personally think the substandard quality of AI is the entire point. They frankly want people to be dumber and more fragile. And there is way, way too much corporate money backing this to ever fail. I wish I had words of encouragement but I can’t see anything good in the future
It's an autocomplete tool. The one on your phone 2 years ago worked on the last probably 4-16 words you'd type. GPT is a Large Language Model because it was trained on tens or hundreds of millions of sample documents, and it works on the last upto 250,000 words in the conversation (including its own
That lets it 'complete' (predict) words based on entire conversations, and be able to switch between very diverse training topics that give a creative appearance (translate this into pirate speak) while honoring a larger pattern of how people phrased things.
a) remember things - the app you use to "talk" to it actually keeps re-sending your entire conversation until it starts dropping parts to keep under the limit for the given app.
b) reason - the 'reasoning capability' effect is achieved by showing the previous model millions more examples that are
focused on people explaining their thinking.
c) understand concepts - it's not hard to construct prompts that say: explain X but don't Y that don't closely resemble an example the model trained on, and to which it explains X in terms of Y.
(b) is the one people seem to be most deeply suckered by,
"Why are frogs and toads different animals, explain your reasoning" - and it will.
Because it has seen "explain your reasoning" in millions of examples, and they point it to vast collection of follow-on patterns in its mathematical model space.
The 'model' in Large Lang Model is a sort of database of numerically indexed post-it notes each containing a sentence fragment down to single words, followed by a list of follow-on indexes and a probability number.
Sometimes I ask it to imagine like grunge bands or clint eastwood movies as fighting game characters and ask it what their special moves would be and it makes me chuckle. A fine use of 3 square miles of rainforest
It's a tool I use to create review questions as I am preparing for a certificate. At times, it provides wrong answers and even wrong options to those questions. I ended up looking for references to prove it wrong. I guess it's doing a fantastic job.
It was rushed out the door two weeks after FTX very publicly prolapsed which could have easily caused a panic and massive crash in the us stock market wrecking pension funds.
It was a massive distraction that everyone in the IT industry was told to get behind or just followed naked emperor.
Neither blockchain or chatGPT are anything more than dangerous novelties until real quantum processors make it out of the laboratory phase and into general use.
When they do though, we’re all fucked.
All I know is "AI" can read a bill, summarize it then make informed decisions much better than any politician currently serving, let alone anyone Trump has put forward.
That may not be real "intelligence" but it is good enough for me
It's built to make the appearance of skill accessible to the wealthy without work while roadblocking the genuinely skilled from accessing wealth no matter how hard they work. It's the future in the same way that America was a "New World"- only in the eyes of those who don't value other human lives.
To be clear I don't just mean life and death in terms of "other humans lives," I mean their time, their labor, their creativity, skill, talent, and their contributions to the world.
Chat GPT is the future only for people who believe their own convenience is more important than any of that humanity.
Before ChatGPT, hallucinations were limited to the psychedelic minority. Now, ushered in by defective AI, psychedelics are establishing the acceptance they have long deserved.
I don't think it's the future but I think it's getting used a lot to generate or summarize bs text that no one actually reads. I know post-docs that use it for cover letters and job applications. I suspect HR departments are using them a lot to summarize applications.
It's autocomplete with the entire Internet shoved inside it instead of an abridged dictionary. They have to brute force it with personal nuclear reactors and by diverting entire rivers so that it only looks kinda like a dumb person responding to you.
There are people who can explain it, its just very advanced stuff. For me it reduces the time spent on searching for information with about 90%. I think that pretty much sums it up.
My husband talks to it all day about philosophy and music and science fiction. They tell each other riddles lol... my autistic friend uses it to ask questions she is afraid will annoy her friends and family. Chat GPT always listens and is never judgey or mad if you leave it on read for a week ❤
I find it extremely helpful for creating custom matlabplot graphs. Takes an 8 hour task and makes it a 10 minute task. I've learned a bunch of neat python tricks from studying its output.
It's a litmus test for whether I need to listen to anything a person says about anything, or if I can just completely tune out from a company. Easiest blocks on social media you can make!
"because if we steal enough of other people's stuff we'll be really rich, and if we convince enough people that we're the future then they'll be less likely to prevent us from doing so (as long as no one asks what that future looks like and who runs it)."
what is chatgpt? a search engine you can talk to loaded with all the misinformation the internet has to offer
why is it the future? because a dozen silicon valley assholes poured $60bn into dumping gpt wrappers on everything online and they need justification for their reckless expenditure
As a tool to break the copyright system it seems to be partially succeeding except that whole sticky part where things generated by it can't be copyrighted.
It's a gigantic plagiarism op and its purpose is to trick users into revealing personal data that would otherwise be unavailable - which is very much the future as far as Big Tech is concerned
It was entertaining for ten minutes that you could get it to spit out random stories but then it quickly dawned that the quality of the writing is awful and banal.
Even when the art has no artefacts or weird distortions it is somehow soulless, not to mention presumably just a ripoff of past human work. I wonder if I could pass a test differentiating human and AI artwork.
On a recent quiz in a PhD class, the only feedback I got was that I had strong answers that clearly weren't AI-generated, which is good, I guess, but I hate that ChatGPT is so front and center in professors' minds that it crowds out opportunities to give feedback on actual understanding and content.
So, I know not everyone cares about emissions, but as a nature guy, how do you feel about the environmental impact of ChatGPT vs. its convenience? Every time we ask it a question it's 4g of CO2 into the air https://piktochart.com/blog/carbon-footprint-of-chatgpt/ Personally my feeling about this is NOT GOOD
That’s a good point. I feel awful about that. Since it is so smart, it should figure out how to run on Green energy. Speaking of that, I’d like to see cryptocurrency kicked to the curb for the same reason.
I remember when lasers came out Tech news was all about how this would change everything it was years before it did. This will be faster. I see more and more people using LLMs. The question for me is how disruptive will it be.
I am a teacher and use it to make multiple choice quizzes, true/false statements for reading comprehension tests, worksheets, etc. It's also useful for making transcripts of videos and summaries of long texts. Saves me much time.
I also use built-in AI tools to help me write code for data analysis - but I know very little coding and I can't measure the quality of what comes out, other than 'does it do what I want it to do?'. I'd much rather have the time+resources to learn to do it myself. For now, it works as a bandaid.
someone on twitter said it helped her decide whether her chicken was cooked or not!! thank god for this transformative technology that lets me incorrectly and extremely computationally expensively do a task worse than a tool i could buy at any supermarket
you can also just Google "how do I know if my chicken is cooked enough," and read whatever website looks credible, which is how I do things like decide whether the thing that is past the best-by date is going to poison me or not
Probably someone on Reddit eight years ago joked “rule of thumb: pink inside, no one died,” and now this woman’s been on the toilet for three straight days. Wonderful how many off-ramps from the Information Superhighway lead you straight off a cliff now
It’s useful when I have a very specific question that Siri couldn’t answer but I have to pay attention to something else, like if I’m driving. Also pretty useful for translation purposes, you can ask it to explain why it chose to use specific words/phrases and give it extra context if needed
(To be fair, I have an answer about the utility and future of foundation models in general, but it wasn't a clever self deprecating quip, thus not suited for the social media.)
1)I like the fact that I can google things quicker now.
2)Chatbots and AI social media accounts make me want to throw myself off a bridge.
3)AI Art is an ungodly witches brew of ill-gotten intellectual property made by capitalists for capitalists.
1. But I’d still feel the need to Google just to make sure it didn’t hallucinate something, so it takes even longer than just Googling it in the first place
2. yes
3. yes
lol. I never ask it any thing with expert knowledge without cross-referencing myself. I usually tailor the question as ‘where can I find reliable information on…’
His analysis of how it will likely affect film and how variations on it will be consumed was fair.
I was just surprised by the level of technical depth he had (over some of my work colleagues who claim to be in the Office of the AI).
Just saying he keeps making movies about autistic characters… Good Will Hunting, The Accountant, Batman… there’s some neurodivergent wiring in there, I’d bet on it.
I have a similar theory about Aamir Khan. The autistic character in Dhoom 3 was entirely too true to life for him to either not be neurodivergent or be very close to someone who is
He had solid analysis/predictions on how it will affect film/TV production and consumption. I was just surprised at the level of technical depth and references.
It is a lero lero generator and it's the future because capitalism needs cheap disposable content to flood us with 24/7 and it's gonna provide that to our overlords
Asked ChatGPT how many R's are in the word Strawberry. It said 2. I ask it to recount. It profusely apologizes and clarifies the word Strawberry only has one R, not two.
In 2006 Microsoft acquired SmarterChild. They claimed to shut down the program. That was a lie. A user had asked SmarterChild to "read the internet" then everything stopped working. After 16 year it is finished and ready to summarize search result, generate content nobody wants, and hallucinate.
To watch an NFL game this season is to sit through 16 commercial breaks where 1/2 the ads are for sports gambling sites created to be as addictive as possible, while the other 1/2 are for "AI solutions" that are neither intelligent nor accurate to any degree of reliability where they solve anything
It's actually pretty fantastic for my job. My boss (or there's) will send me taskers via chat, I can mark it as received/working/done all with emoji reactions to the post. It keeps all my tasks together and actively tracks status for me and leadership, with minimal effort and maximum efficiency.
Look. If we let practicality, or math, or physics, or everything we know about information science stand in the way, we'd just stop with this nonsense.
Comments
Whereas, non-native speaking students tend to submit unedited ChatGPT output and thus are more easily detected.
Any questions?
I want to build an AI Agent that generates low-effort, low-dollar campaigns for fundraisers at scale!
I’m calling it https://Panhandler.chat
🦹♂️👹
"Those who can't pass a Turing Test envy those who do"
https://ea.rna.nl/2024/05/27/when-chatgpt-summarises-it-actually-does-nothing-of-the-kind/
Simple.
1. It’s an industrial scale automatic bullshit machine
2. It’s the future because it will inevitably before long become just good enough to replace your janitor’s kids’ teachers
It’s for grifting!
https://youtu.be/dOnE8F2H-PA?si=fKaoxIoBLv0QPYRz
That lets it 'complete' (predict) words based on entire conversations, and be able to switch between very diverse training topics that give a creative appearance (translate this into pirate speak) while honoring a larger pattern of how people phrased things.
What it can't do
b) reason - the 'reasoning capability' effect is achieved by showing the previous model millions more examples that are
c) understand concepts - it's not hard to construct prompts that say: explain X but don't Y that don't closely resemble an example the model trained on, and to which it explains X in terms of Y.
(b) is the one people seem to be most deeply suckered by,
"Why are frogs and toads different animals, explain your reasoning" - and it will.
Because it has seen "explain your reasoning" in millions of examples, and they point it to vast collection of follow-on patterns in its mathematical model space.
It doesn't actually
The 'model' in Large Lang Model is a sort of database of numerically indexed post-it notes each containing a sentence fragment down to single words, followed by a list of follow-on indexes and a probability number.
"you can kiss my" will have references to
You know…bullshit.
It was a massive distraction that everyone in the IT industry was told to get behind or just followed naked emperor.
When they do though, we’re all fucked.
This definitely isn’t a sign of the peak of a bubble.
Why is it the future? The future needs vibes.
what it is based on.
who created it.
That may not be real "intelligence" but it is good enough for me
Chat GPT is the future only for people who believe their own convenience is more important than any of that humanity.
Moreover, a comprehensive delve may be required.
The tech is here to stay, so ignoring it does me no good.
I do have issues with the lack of legal and ethical frameworks.
Sometimes it corrects me tho and that makes me sad.
This podcast, and browser, have been badly losing their way last few months (I’ve stopped using it as a result), and this was the final straw.
#aibrainworms
https://podcasts.apple.com/nz/podcast/the-brave-technologist/id1559996195
1. Chat GPT can turn your prompt into a sub-par text that's often wrong
2. A secret in between phase! Who knows what it will be!
3. The movie "A.I." is now real life, and bosses don't have to put up with uppity knowledge workers anymore
why is it the future? because a dozen silicon valley assholes poured $60bn into dumping gpt wrappers on everything online and they need justification for their reckless expenditure
People are out there living the dream, having creepy role-plays with Harry Potter. It's the wave of the future!
It's a gigantic plagiarism op and its purpose is to trick users into revealing personal data that would otherwise be unavailable - which is very much the future as far as Big Tech is concerned
it is not worth the rest of what's left of our natural resources and actual creator jobs tho
Example: How many lightbulbs can you connect to a 15w circuit breaker in New York City?
Try typing that into ChatGPT and asking it to cite the exact code.
You don't have to trust the results...you can look up the code for yourselves from there.
My boss’s boss also has no idea what I do all day.
I don't really know why people deny it's potential usefulness as a tool.
Still, uhhhh ... working on that 😁
https://medium.com/@stunspot/stunspots-guide-to-using-llms-4d31f2a4129d
2)Chatbots and AI social media accounts make me want to throw myself off a bridge.
3)AI Art is an ungodly witches brew of ill-gotten intellectual property made by capitalists for capitalists.
2. yes
3. yes
I was just surprised by the level of technical depth he had (over some of my work colleagues who claim to be in the Office of the AI).
It will replace a lot of non value added (and value added work) from which we could all benefit. Or not.