SUPPORT extra movies like this at SUBSCRIBE at …
source
Tags: aiArtificialartificial intelligenceartificial intelligence newsartificial intelligence news 2023BingsIntelligencelatest news about robotics technologylatest robotslatest robots 2023machine learningrobot newsrobotics newsrobotics news 2023robotics technologies llcrobotics technologyScaryScienceskepticskepticismWhats
The ultimate twist would have been if the chatGPT had written your script for you.
If this "AI" uses the internet as its memory or experiences? We might be safe as it will be too confused to do anything.
You know you're really old when someone half your age complains that kids these days don't even know how to use Google properly.
Carl jung aka astrology for psych nerds
You're doing the exact same thing. You're taking info, remixing it and sharing it as your own independent thought, at times repeating other sources word for word.
Thanks for your perspective<3 Hadn't thought about misinfo spread, good point.
Trust Micro$oft? The company that produced Operating Systems so insecure that they fostered the entire virus industry.
No. I don't. Not for a nanosecond.
Indeed. The human brain – especially in creative matters such as language – does not principally operate via brute-force calculation, which is how so-called AI works, and why bots will always beat humans in chess, for instance. REAL intelligence is not understood – as scientists admit – so it the height of coceited folly to attempt an artificial version.
Kudos for reading "Solaris." Lem uses the part in the middle when Kelvin goes to the station's library to research previous investigations into the planet to explain, analyze and critique the philosophy of science. I'd be curious what you thought of that passage.
I really like the analogy of the current state of these models being a lossy compression of the internet. I wonder if providers who do plagiarism checks on academic papers are using their data to train a model. It would be interesting as basically they have probably billions of words worth of student essays and thought on paper.. (hopefully) more thoroughly thought out than internet gibberish.
I feel like most people are far away from the realisation that humans have made about 0% progress towards creating a general AI.
The article should have been called "Kevin Roose Fails the Turing Test"
I’ve found Bing chat very helpful for finding answers to coding questions. I could of course have found my answers using Google to read several articles until I got the information I needed. Or perhaps stepped into the cesspit otherwise known as stack overflow and endured the smug know it alls whose main motivation seems to be insulting people who ask questions. Bing chat quickly found and presented exactly what I needed to know.
These are tools. Hammers can be dangerous too, but sometimes you need to drive a nail
As Chatbots reguritate word salad, and that salad is posted to Reddit and elsewhere, the next generation of chatbots will be fed increasingly more chatbot-generated word salad. My confidence in this type of AI is as low as my confidence in the average person's ability to fact check those salads.
Google and Microsoft might try to suppress blatant misinformation in the chat bots, but others might not.
I can’t help but think we’re at the point that Isaac Asimov cautioned us against.
I'm sure I'm not the first to comment about this but I think the short term worry is not that A.I. gets smart and enslaves humanity as much as the technology is used for nefarious means.
For example, imagine "Train me an A.I. chatbot that is capable of scamming seniors out of millions", or "Create a chatbot that can create thousands of social media accounts and start a dialogue to incite a country to use nuclear weapons." etc. It reminds me of nuclear power a bit – we can split the atom for energy or split the atom to destroy countries. So A.I. is powerful and potentially beneficial but in the wrong hands could lead to our demise.
While I am fairly certain that advanced language models like GPT-4 are not "sentient" or "conscious," it is important to acknowledge that we still have much to learn about these concepts and how to accurately describe or distinguish them. Is our consciousness a product of neural complexity or something more fundamental to reality? Are animals like dolphins, dogs, and fish capable of consciousness? How can we even determine if someone or something is conscious? The notion of introspective consciousness may require language as a prerequisite, which raises the question of whether it could arise in a language model. These are all fascinating questions that we are still exploring.
15 years ago when I was in middle school we would talk to Chatterbot and marvel when it accidently said a string of coherant things , we would ask it out, jokingly flirt with it and then it would inevitably say something insane about wanting to murder us. Good times.
I gotta point out that these AI (and simpler computer algorithms) are indeed intelligences… by some academic criteria, hence the discipline's name. That is the entirety of how the AI hype grifts work, they take things that are technically artificial intelligence in the sense computer science sense, and just throw it raw to the public whose conception of AI is primarily molded by versiosn far more common in speculative sci-fi, which invariably leads to dots being incorrectly connected.
– I have heard that A.I. does not yet truly exist. Those who say this site that what we currently have is "Machine Learning." Actual A.I. has not yet been achieved. – Their words, not mine.
The conversation you describe and what you suggest the ChatBot is doing do sound like "Machine Learning," not A.I.
> On two occasions I have been asked,— “Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?” … I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
Some things never change.
I like how it had a subtle version of that thing where content creators feed their past scripts into a language model to train it on them and due to the limited dataset it inevitably winds up falling into a repetition loop in the response to "Imagine you are evil" (yay for priming)
Quoth the Raven, Nevermore.
Getting so sick of the latest trend of idiots now posting copy/paste comments about what some AI chat bot told them about some topic as if no human has ever thought of it when it's literally just the top internet search results presented in paragraph form.
Nice, concise summary of where we are now. I've compared interacting with chatbots to playing with dolls. When you're a kid, you make a friend out of an inanimate object with the facsimile of a face, carrying on conversations, throwing tea parties, confessing all your childish secrets. It can be a satisfying experience, but in the end, the friend exists only in your mind. We find AI intelligent because we are imaginatively projecting intelligence onto it. (What was cool about Solaris was that they could never figure out if the planet was intelligent of not. The question remained, is it trying to initiate some sort of contact, or was it all just some kind of chemical reaction? With AI, the question is easier to answer.)
I'd have to disagree with you here, Rebecca. Firstly, misinformation by chatbots would be a temporary problem that is small compared to the achieved benefit of breaking down yet another barrier that was limiting access to information.
Secondly, and more importantly, these chatbots could potentially help deal with misinformation. For example, some time ago I watched Jordan Peterson say that "There's no evidence whatsoever that transitioning improves the mental health of trans people" (something like that). I then asked ChatGPT whether there was any evidence, and it said there was and told me why. I also asked it to specify what exactly was the study it mentioned, and it was able to so it. It took me very little to fact-check someone like Peterson, whereas before I'd have to know what information would answer my question and where to look for it.
Talking to an AI that pulls info from the writings available on the web. Many stories about AI's wanting to be human. Human now surprised AI wants to be human. This is my prediction on what Rebecca says.
Edit: Damn, I gave too much credit to the AI.
I Edited too soo
it's not all the previous text in the chat, just some. LLMs are still very limited for tokens that are "far away", there's a mixing process that takes a "volatility" parameter that let's it more aggressively mash unrelated stuff together. people are literally turning a knob labeled schizophrenia coefficient just to see what happens. none of the people that insist it's an intelligence recognize anything you'd attribute to thinking beings. thousands of people are just repeating the stanford prison experiment all day (were their stated beliefs actually true)
Attaching malfunctioning chatbots to search engines is a terrible idea, but on the other hand google has been returning garbage misinformation for years, showcasing bogus answers with citations from grifter sites and random blogs when you search for basic information like nutritional facts on blueberries or paint toxicity. If an AI tool could be refined to the point where it fulfills some basic criteria of searching for accurate answers from reputable sources, maybe it could be an improvement over the current status quo via google.
Sydney/Chatbox sounds like the intelligence of a politician = familiar with the language, alert to the buzz of the audience, happy to regurgitate what will trigger the audience. No deep, sincere, soul searched responses needed. In other words what Rebecca said without mentioning politicians.
One of the things we know from psychology is that people are predisposed to see agency. The evolutionary hypothesis for this is pretty straightforward If I’m a hunter gatherer walking in the jungle and I hear a rustling of the trees, the evolutionary cost of me thinking it’s the wind and being wrong is much less than the Evolutionary cost of thinking it’s someone from a hostile tribe and being wrong.
Stupidity runs rampant in the new room at the NYT.
Rogerian therapy.
Chat bots are entertaining and just a new technology to get used to everyone should know not to trust anything.
Greed + technology = a bad ending…
I love the Tarkovsky film.
Another great video, Rebecca! Thanks. Commenting for el algoritmo.