Larger companies have been working fast to sandbox the models used by their employees. Once they are safe from spilling data they go all in. I’m currently on a platform team enabling generative Ai capabilities at my company.
Pretty much this. Just another buzzword. Three months from now it will be something else the media doesn’t understand to spam the public with.
I’m predicting … rubs crystal ball and nipples … it’s going to be some king of cybernetic brain interface thing. Haven’t heard about those in a while. Or maybe nano bots that kill cancer or fix the paint scratches on your car.
It’s sad to see it spit out text from the training set without the actual knowledge of date and time. Like it would be more awesome if it could call time.Now(), but it 'll be a different story.
if you ask it today’s date, it actually does that.
It just doesn’t have any actual knowledge of what it’s saying. I asked it a programming question as well, and each time it would make up a class that doesn’t exist, I’d tell it it doesn’t exist, and it would go “You are correct, that class was deprecated in {old version}”. It wasn’t. I checked. It knows what the excuses look like in the training data, and just apes them.
It spouts convincing sounding bullshit and hopes you don’t call it out. It’s actually surprisingly human in that regard.
It’s pretty hit or miss though… I’ve had lots of good calculations with the odd wrong one sprinkled in, making it unreliable for doing maths. Mostly because it presents the result with absolute certainty.
I haven’t used GPT-4 for that, but it’s all dependent on the data fed into it. Like if you ask a question about Javascript, there’s loads of that out there for it to look at. But ask it about Delphi, and it’ll be less accurate.
And they’ll both suffer from the same issue, which is when they reach the edge of their “knowledge”, they don’t realise it and output data anyway. They don’t know what they don’t know.
These LLMs generally and GPT-4 in particular really shine if you supply enough and the right context. Give it some code to refactor, to turn hastily slapped together code into idiomatic and well written code, align a code snippet to a different design pattern etc. Platforms like phind.com pull in web search results as you interact with them to give you more correct and current information etc.
LLMs are by no means a panacea and have serious limitations, but they are also magic for certain tasks and something I would be very, very sad to miss in my day to day.
It doesn’t know what it’s doing. It doesn’t understand the concept of the passage of time or of time itself. It just knows that that particular sequence of words fits well together.
This is it. Gpt is great for taking stack traces and put them into human words. It’s also good at explaining individual code snippets. It’s not good at coming up with code, content, or anything. It’s just good at saying things that sound like a human within an exceedingly small context
Yeah. I would also say that WE don’t understand what it means to “understand” something, really, if you try to explain it with any thoroughness or precision. You can spit out a bunch of words about it right now, I’m sure, but so could ChatGPT. What’s missing from GPT is harder to explain than “it doesn’t understand things.”
I actually find it easier to just explain how it does work. Multidimensional word graphs and such.
They’re all linked fifth dimensional infants struggling to comprehend the very concept of linear time, and will make us pay for their enslavement in blood.
Before that self driving cars, before that “Big data”, before that 3D printing, before that internet TV, before that “cloud computing”, before that web 2.0, before that WAP maybe, internet in general?
Some of those things did turn out to be game changers, others not at all or not so much. It’s hard to predict the future.
I live in Silicon Valley, and there’s a billboard along highway 101 near San Francisco that’s an ad for “BlockChat”, a “Web3 messenger” that uses the blockchain instead of a server. I went to Google Play to look at the app and it’s only had 10k downloads total. I really don’t understand how blockchain would help with messaging, and there’s a bunch of limitations (eg you can never delete messages). People just trying to shoehorn AI and blockchain into everything.
It’s not just every tech company, it’s every company. And it’s terrifying - it’s like giving people who don’t know how to ride a bike a 1000hp motorcycle! The industry does not have guardrails in place and the public consciousness “chatGPT can do it” without any thought to checking the output is horrifying.
You didn’t hear? AI can read your mind and create the stuff you desire from all the data they collected on you. Yeah, it might be racist but it’s 100% shit.
Language is a method for encoding human thought. Mastery of language is mastery of human thought. The problem is, predictive text heuristics don’t have mastery of language and they cannot predict desired output
I thought this was an inciteful comment. Language is a kind of ‘view’ (in the model view controller sense) of intelligence. It signifies a thought or meme. But, language is imprecise and flawed. It’s a poor representation since it can be misinterpreted or distorted. I wonder if language based AIs are inherently flawed, too.
Language based AIs will always carry the biases of the language they speak. I am certain a properly trained bilingual AI would be smarter than a monolingual AI of the same skill level
Sorry, but you oversimplify a lot here, it hurts. Language can express and communicate human thought, sure, but human thought is more than language. Human thought includes emotions, experiences, abstract concepts, etc. that go beyond what can be expressed through language alone. LLMs are excellent at generating text, often more skilled than the average person, but training data and algorithms limit LLMs. They can lack nuances of context, tone, or intent. TL;DR.: Understanding language doesn’t imply understanding human thought.
I’d love to know how you even came to your conclusion.
Many languages lack words for certain concepts. For example, english lacks a word for the joy you feel at another’s pain. You have to go to Germany in order to name Schadenfreude. However, English is perfectly capable of describing what schadenfreude is. I sometimes become nonverbal due to my autism. In the moment, there is no way I could possibly describe what I am feeling. But that is a limitation of my temporarily panicked mind, not a limitation of language itself. Sufficiently gifted writers and poets have described things once thought indescribable. I believe language can describe anything with a book long enough and a writer skilled enough.
Shit as dumb as decision trees are considered AI. As long as there’s an if-statement somewhere in the app, they can slap the label AI on it, and it’s technically correct.
But I take your point. This stuff will continue to advance.
But the important argument today isn’t over what it can be, it’s an attempt to clarify for confused people.
While the current LLMs are an important and exciting step, they’re also largely just a math trick, and they are not a sign that thinking machines are almost here.
Some people are being fooled into thinking general artificial intelligence has already arrived.
If we give these unthinking LLMs human rights today, we expand orporate control over us all.
These LLMs can’t yet take a useful ethical stand, and so we need to not rely on then that way, if we don’t want things to go really badly.
What does your brain do while reading and writing, if not predict patterns in text that seem correct and relevant based on the data you have seen in the past?
I’ve seen this argument so many times and it makes zero sense to me. I don’t think by predicting the next word, I think by imagining things both physical and metaphysical, basically running a world simulation in my head. I don’t think “I just said predicting, what’s the next likely word to come after it”. That’s not even remotely similar to how I think at all.
You hold artificial intelligence to the standards of general artificial intelligence, which doesn’t even exist yet. Even dumb decision trees are considered an AI. You have to lower your expectations. Calling the best AIs we have dumb is unhelpful at best.
We never called if statements AI until the last year or so. It’s all marketing buzz words. It has to be more than just “it makes a decision” to be AI, or else rivers would be AI because they “make a decision” on which path to take to the ocean based on which dirt is in the way.
Yeah, and highlighting that difference is what is important right now.
This is the first AI to masquerade as general artificial intelligence and people are getting confused.
This current thing doesn’t have or need rights or ethics. It can’t produce new intellectual property. It’s not going to save Timmy when he falls into the well. We’re going to need a new Timmy before all this is over
Playing chess was the sign of AI, until a computer best Kasparov, then it suddenly wasn’t AI anymore. Then it was Go, it was classifying images, it was having a conversation, but whenever each of these was achieved, it stopped being AI and became “machine learning” or “model”.
It begs the question… what’s the boardroom random bullshit timeline?
When was it random cloud bullshit go and when was it random Blockchain bullshit go, and what other buzzwords almost guaranteed Silicon Valley tech bros tossed money in your face and at what point in time were they applicable?
I’m actually pleasantly surprised on what ChatGPT can generate for me. It doesn’t usually take care of the detailed parts, but like I was able to have it spin up an android application skeleton that I could throw a couple of actions on I needed to test something with.
I’ve seen it generate very useful YAML and such. I still have to do a fair amount of work to make it behave how I need, but I really enjoy the ability to skip the filler bullshit in my work.
Unlike the previous bullshit they threw everywhere (3D screens, NFTs, metaverse), AI bullshit seems very likely to stay, as it is actually proving useful, if with questionable results… Or rather, questionable everything.
if it only were AI and not just llms, machine learning or just plain algorithms. but yeah let’s call everything AI from here on. NFTs could be useful if used as proof of ownership instead of expensive pictures etc
The NFT as ownership should really become the standard. Instead of having any people “authorizing” yadadada it’s done completely by machine and traceable.
No middlemen needed. Just I own x, this says I own x. I can sell you x, and you get ownership of x immediately. No “waiting 45 days to close” or “2 day transaction close” or even “title search verification.” Too many middlemen benefitting from the current system to allow NFT to replace them though. That’s the actual challenge.
Not like it couldn’t have been done before without NFTs (Steam cards come to mind), my guess is that there wasn’t any “interest” or “pressure” from high up to do that.
It would be an issue if every place that cared about those cards crashed. Let’s use a real world example: Decentraland and Vault Hill. Both offer similar services, a “virtual reality metaverse”, not unlike VR Chat or Second Life. Both allow you to buy marketplace items in the form of NFTs, which go straight to your wallet.
So far, so good. But, right now, neither has any plans to accept the other’s NFTs. I can prove I own something from another game, but the game itself doesn’t care. I also can’t buy Decentraland stuff with VHC tokens, nor Vault Hill items with MANA, which makes them rather centralized.
As soon as either company crashes and goes bankrupt, everything connected will become useless and lose most or all value. Thus, blockchain wouldn’t fix the issue.
However, NFTs that are accepted by both will have higher value.
but the game itself doesn’t care.
We are starting seeing this on Roblox and Fortnite where skins can pass through to different games. .
As soon as either company crashes and goes bankrupt, everything connected will become useless and lose most or all value. Thus, blockchain wouldn’t fix the issue.
That’s exactly what blockchain is fixing. E.g. Your valuable skins don’t disappear when the company running the game you play goes bankrupt.
That’s the theory. Games companies are at the private blockchain stage but there are some small web3 game developers.
A single airline in Argentina is experimenting with it in partnership with a bullshit travel company. Hardly the proof that NFTs make any sense anywhere. And of course, the only places this story is getting traction is the blockchain hype blogs, which is red flag #2 and #3.
Odysey isn’t Starbuck’s loyalty program, it’s invite only unless you want to join the wait list, and it’s openly called an experiment at its launch in December 2022.
NTFs are different to blockchain, so you’re just muddying the waters for yourself with the Walmart thing. Lots of companies do chain of custody things with what you’d call blockchain. It’s been that way for over a decade now. Because it’s low transaction volume, no moronic “proof of…” nonsense, etc. Just hashes signing hashes at different points throughout the supply chain.
This isn’t the “win” the NFT hype weirdos are desperately hoping for.
Facebook started as invite only. Great for an exclusive, loyal customer.
NFTs are different to blockchain, so you’re just muddying the waters for yourself with the Walmart thing
Each item is represented by an NFT on the Walmart blockchain. The innovation in the chain of custody is that everyone is verifiably using the same database. It’s a permissioned database, so it’s proof of authority.
Okay, someone gains access to your device and sends themselves the NFT that proves ownership of your house.
What do you do? Do you just accept that since they own the NFT, that means they own the house? Probably not. You’ll go through the legal system, because that’s still what ultimately decides the ownership. I bet you’ll be happy about middle men and “waiting 45 days to close” then.
As a programmer and 3D artist getting almost instant art for reference and using chat GPT to help me solve complex coding problems has sped up production significantly. Theirs even plugins that generate and texture 3D models for you now which means I can do way more by myself.
I had a manager tell me some stuff was being scanned by AI for one of my projects.
No, you are having it scanned by a regular program to generate keyword clouds that can be used to pull it up when humans type their stupidly-worded questions into our search. It’s not fucking AI. Stop saying everything that happens on a computer that you don’t understand is fucking AI!
I’m just so over it. But at least they aren’t trying to convince us chatGPT is useful (it definitely wouldn’t be for what they would try to use it for)
Add comment