Yes they announced Bing has been Chat GPT 4 since going on waitlist publicly
OpenAI Introduces GPT-4
- Thread starter wingkongex
- Start date
This is a bit mindblowing. Now if Apple could just plug this into Siri, maybe we could actually have a true digital assistant like Apple's 1990s vision, that can do all sorts of mundane internet tasks autonomously.
I started wondering to myself how the stock market would react if there were hyper intelligence AIs instantly buying and selling (more-so than the current stock algorithms) based on imperceptible metrics.
Reminds me of MGS2 and The Patriots.
Yea, I don't think i'll be using my licenses in a decade or less.
A lot from my initial testing. Here's me trying to do the AI dungeon thing.
A lot of folks in my social bubble who got into it & dev jobs are starting to be legit scared for their future.
We'll see a lot of fundamental changes in the next years.
And while I appreciate the optimistic idea of establishing entirely new jobs around the changing realities of work, I often feel like that sentiment disregards the likely reality of AI/automatization "improving" at such a rate that the new jobs are gonna be useless within a short amount of time again
We'll see a lot of fundamental changes in the next years.
And while I appreciate the optimistic idea of establishing entirely new jobs around the changing realities of work, I often feel like that sentiment disregards the likely reality of AI/automatization "improving" at such a rate that the new jobs are gonna be useless within a short amount of time again
How far away are we from Chat-GPT figuring out how to more efficiently improve itself, when prompted?
The link is the neverending race to the bottom inherent to capitalism, the race to cut business expenses to the absolute bare minimum. One of that race's biggest hurdles is the need to pay human beings a wage to live on.
Artificial intelligence research and development isn't being done for its own sake, for the sake of science, for the promise of a post-scarcity world, or for the fun of it. AI R&D receives funding because there is future money to be made, and that money comes from selling that tech to businesses who will use it to cut as much human labor from their workforce as the technology can reasonably allow them to. The expectation is that as the tech advances, so to will the tech's capacity to displace human workers.
It is.
Confirmed: the new Bing runs on OpenAI’s GPT-4
Congratulations to our partners at Open AI for their release of GPT-4 today. We are happy to confirm that the new Bing is running on GPT-4, which we’ve customized for search. If you’ve used the new Bing preview at any time in the last five weeks, you’ve already experienced an early version of...
blogs.bing.com
Yeah, I think it's really the rate at which it's evolving is what's scaring people. Choosing a major in college will be even harder because who knows if the jobs for that field will be worth anything by the time you're done with your degree. It's crazy to think what it's gonna be like a year from now. Exciting but terrifying at the same time.
I think its pretty useful for 'commodity' stuff; I mean: if you can phrase the right question and find an answer on Stack Overflow, I'm not that surprised that AI can generate the corresponding code if you put in the right prompts, even if its a bit wonky. I do wonder how it performs when you get into more complex stuff (for example, modern 3d rendering), but since I'm not familiar with that I can't really judge it.
The kumbaya dreamscape where corporations pay similar wages to people for less work because they're going to make it up and then some through generative AI is just that, a dream. It's going to be very ugly and very bloody before it gets to where we ideally want to be, if ever.
What's scary to me is the people at the "top" are just as clueless as the rest of us. They're confused and rushing to catch up and with that will come further messiness. This is orders of magnitude more impactful than cryptocurrency but regulation is moving just as slowly. Everyone is on their back foot because it's advancing way too quickly for us to deal it.
Nah. Technological advancements like AI are a good thing. That Governments have refused to do their fucking job and regulate isn't on the people making them. Properly regulated and used it will lead to a new renaissance for humanity.
Experts have been screaming from the rooftops the threat AI poses to society due to how it's configured for decades now. And the political class decided to sit on their asses and do nothing about it while the population just ignored then.
Well, this is the consequences of ignoring the warnings. We've seen tech improve year over year. It was only a matter of time.
Speaking as a dev... Anyone who is seriously scared for their jobs obviously hasn't been tasked with decoding what their clients want. AI is good, but it's not at the level where it can parse a week's worth of 2+ hour meetings of a client speaking bullshit and using words wrong to request a feature then getting upset because they wanted the complete opposite based off of some obscure rule they never described in detail, but you had to infer out of thin air.
I have a lot of misgivings about AI but you kind of can't ignore it. Learning how to work with it is going to be huge not just for so many industries, but just for almost anything you want to do creatively or to make money independently. The potential of this technology is truly quite remarkable, even if I expect lots of people to be hurt by the upheaval it will cause. It does nobody any favors to turn away from what the future is going to be.
yeah if you don't see where this is heading then you really do have blinders on, regulation is going to be slow and it's going to become far more messy before it becomes better.
It's not like leadership are aware of this either though, some of the companies I've worked with are only really just starting to learn why offshoring was such a bad idea after 20 odd years of doing it
It's not like leadership are aware of this either though, some of the companies I've worked with are only really just starting to learn why offshoring was such a bad idea after 20 odd years of doing it
It's a mistake a lot of people are making. They hate it, very understandably, but they also want it to disappear, which it just plainly won't. The exact opposite, in fact. It's going to become very ubiquitous very quickly.
At this point we have to learn to work with it. Learn how to prompt it well to get what you want. That's going to become a new language of sorts, like google-fu but 10x.
You do yourself a disservice by ignoring it even if you have a genuine (and understandable) ethical quarrel with it.
yeaaahhh.. hahaha as if that will happen anytime soon. Listen this will turn into a huge unregulated mess before it will be forced to sort itself out.
Changes happen too fast that before regulatory bodies have wrapped their heads around one model, 10 new ones will have come out.
We still have governments comprised of politicians who barely know what the internet is.
I have zero faith humans using this in ways that are ethical and those who develop it to spend equal amounts of time and money into regulating it as they are in developing it.
There's not even individual authors credited on the "paper", it's just written by "OpenAI".
It's a shame because they actually did use to be open until around 2019 or so. The bullshit phrasing of "competitive landscape" and placing it next to the very real safety concerns is awful, especially because that whole "Reinforcement Learning from Human Feedback" bit is just a euphemism for "we paid a bunch of people to manually moderate and babysit our chatbot so it doesn't say awful/incorrect shit".
It wont just be devs, and It won't be a complete workforce replacement at first. It will be used to gradually cut workforce size little by little by using this tech to increase the productivity of the employees remaining. Why pay a few grand each month for two employees when you can pay one employee's wage and a (relatively) small subscription fee to double that employee's productivity. That person can handle the messy human interpretation aspect you're talking about, and you can bet they won't suddenly be getting twice the salary.
That pattern will continue and be pushed to absolute limit it can handle to the point where you might see one single person in a space where you used to have a team, and that one person's role will be to add the human nuance, to manage the mistakes made by the machines, like the one employee who looks after the half-dozen self checkouts at the grocery store.
I asked ChatGPT3 this and it gave me an exceedingly well written document on how to improve itself...so I'd say we're there.
Exactly. You nailed it really. That is 100% what will happen, particularly that final analogy you gave.
Oh, I completely agree. But there is also only one group of people at fault for that state of things. The voters. If the voters did their fucking jobs and voted correctly, working to get new blood in there instead of dinosaurs one fall away from an early grave, we wouldn't be in that mess. And yes this is overly simplistic, but this isn't a thread on why things are the way they are.
Instead we've devolved the whole process into a sports match between a handful of teams dedicated to trying to elect the oldest state of grifters possible who are all trying to do the least amount of work.
But hey, sometimes the system has to completely break before people realize how fragile it really was.
Last edited:
For sure. But also, you'd be amazed how many companies still operate off of ancient non-digitized paper processes. Even after COVID where you'd think they would have digitized them to keep their businesses running. And when they don't run off paper processes they're using something developed in the 80s that everyone is terrified of looking at least it break.
Every company has a backlog of projects they have been putting off to reduce this, but that backlog grows more than it shrinks. So the threat really isn't there yet.
Keyword yet, because that day will come. And AI will accelerate it. And teams will absolutely shrink rather than being replaced wholesale. But it's not happening yet.
Didn't they just drop the price of the 3.5 api too
Not sure how much more impressive 4 is in practice
I don't know - every time I use it I get less confident in it. I asked it to write some code for me and it literally invented a non-existent API for the framework I was using with full explanations. Confidently incorrect in the worst way possible. It can automate some trivial stuff but to be honest I haven't encountered a situation where it actually saved me time relative to doing it myself.
This really depends on the specific entities doing the research. The is quite a bit of AI research that is being done for the sake of science but the funding for it isn't there compared to the private incentives because well, that would require funding public research institutions like universities. The best answer to that is to actually make sure that there is funding in place for people who are interested in the science so they don't get brain drained away from academia towards the private sector.
There's also the fact that a lot of AI research is being done for military purposes which I think is just as bad if not worse in terms of risks than the strictly for profit use cases because anything done in the name of "national security" will receive much less scrutiny from everyone across society.
Like I said/examples above, from my limited testing it's a lot more descriptive and lengthy with it's responses. The fact it can have 8k or 32k token limit compared to 4k is enough to make it better. 3.5 loses the context of the conversation quickly since it's limited to 4k tokens.
Last edited:
I imagine the prices will probably come down once more competitors start showing up.
At this point the only competitor will be Google, they're too far ahead of the game and the time/costs to build/run something like ChatGPT 4 are massive.
Last edited:
The key is to use Copilot and not just ChatGPT responses, its significantly better within the context of the rest of the code base than just in chat/bing search snippits
Maybe we should get in front of allowing AI to lie to humans rather quickly....
ChatGPT is terrible at math. It talks through the concepts pretty well and mostly describes the right steps, but then it just randomly mixes up variables, or divides when it says it's multiplying, etc.