Warning AI pt. 3 (1 year update)
“AI couldn’t take an analysts job”, “Schizoposting is back now”,”Oh I’m a CS major too and went to a shitty university. AGI is decades away.” Loved your nice comments on my first post and most recent
Just wanted to remind everyone the amount of shit I caught for warning everyone about AI a year ago. I was chicken little until the sky actually fell. Fact of the matter is as much as I despise most of you lot for thinking you’re smarter than you are, you are important and influential. Look Copilot is taking every analysts job. As soon as the cost of compute to run these models goes down, which it will, analysts and associates are gone. The GPT plug-ins in their alpha stage are significantly more capable than most of the excel monkeys on this site. Tell your MDs to shove it for once and stop turning AI into a race so they can get a nice bonus and rent a yacht for a week to pretend they are Bezos. You guys handle deal flow. You guys are the ones who let Google and Microsoft buy out all these AI startups so they can literally not build your replacement, but an actual fucking god.
I hope some of you are starting to realize that we are significantly closer to AGI than even I thought in my first post where you called me a schizo. I said last year, probably a decade away, this year I say, 5-6 years at most. GPT 8 or 9 is probably going to be AGI or if it’s not, some weird chatbot that hallucinates hard enough to believe it is an AGI and actually experiencing consciousness. Guys, I’m sorry for talking shit, but we are seriously fucked. I worked as a quant who specialized in ML and now at an AI company everybody knows. AI makes the nuke look like a joke and unlike the nuke the first AGI will be the last one. The work I do is genuinely terrifying and I would never do it if not for the fact i know we are moving way too fast and alignment of AGI is more important than any political debate going on right now. I’ll be back in another 6 months. See you then.
Deleted
a
It's not even about replacing a few analysts. This technology is history-changing. 2023 will be remembered in entire human history for being the point of AI singularity. Many, many jobs will never be the same.
Update: what’s with the monkey shits ??? Meh, the truth sayer always get hated.
Ah the post of someone who heard some information somewhere but doesn't know shit about how it works.... "embedding GPT" doesn't mean anyting. The search bar isn't even truly embded into windows anymore, it has internet based access. GPT-X is a cloud based LLM that uses pretrained language data to generate responses. The model itself is over a terabyte at this point and needs constant data feeds to remain relevant in a time series basis.
So yes, even if it is "in the desktop" it still has to comply with the insane levels of sercurity compliance regulations the industry has.
Let me put this way: once the pre-trained model and all of its matrix parameters are in the operating system memory , we don't need internet to access the AI. It will occupy large amount of storage and make CPU slower, yes, but Microsoft is exploring ways to make it work, and they will likely achieve it since they have tens of billions of $$$ to burn, thanks to the TMT bankers
the first line is ironic considering how clueless most of your comment is...
Wrong, windows search is integrated directly with the windows shell, it has file system access. internet connectivity is just an added feature.
openai's GPT models are cloud based LLMs. every single other language model, included open sources ones, are also GPT models, and many are not cloud based.
this and the rest of the sentence is so wrong its almost funny. the TRAINING data is likely terabytes, not the model. noone knows how large the "model" would be if stored locally but comparable open-source models are ~300-500gb, and some run on 8gb of ram. some can run on iphones, some on browsers.
first i dont think you understand what a time series is because it makes no sense in this context. second its literally in the name: "generative pre-trained transformer." the model is trained once (it takes months) and after that the model may go through reinforcement feedback processes but the training data is the exact same. current models get up-to-date information by browsing the web or connecting to APIs.
just run a fine-tuned model on an internal server and dont give it internet access. it sounds like you only know about the openai/google etc models, but there are dozens of others.
There’s so many replies on this comment so I’m just going to break down how it works very simply and also explain how it will progress in the years to come from a technical standpoint. GPT and other LLMs used supervised learning to produce its outputs. Essentially all the data that it has ever come across is embedded and thrown into what’s called a vector database (informally referred to as vectorstore) which contains all the information it was trained on. When given a prompt, the model searches thru the vector database to find a series of numbers that is the most resembles what it was prompted. From there it uses statistical methods to find what numbers are most likely to come next translate that into a format you can understand and produce the output.
Like others have said, some things it’s super good and for other things like first order logic which another commentor referenced it’s not that great at. However, this thing is pretty damn good at programming and will continue to progress in its mathematical capabilities. Expect a few more versions of GPT-esque models to come out. All supervised. However, eventually one of those models will become so intelligent (not AGI) that it either convinces itself it is an AGI even though it’s not, or will be used by certain companies I won’t name to actually program the real AGI who’s training will be unsupervised. The problem right now is that these Models are written in very high level languages like C++ & Python so they aren’t as efficient as they need to be for an AGI. The final few versions of models like GPT will be used to program the actual AGI in either Assembly or plain Machine Language.
bump
I've been bullish on your posts because you seemed to have a grasp on the idea that none of us know how crazy this can get.
That being said, being wrong or right about the world ending doesn't matter; the world still ends. I'm an incoming SA and don't know what to do frankly. I feel like everything I've gone down this path for is going to all be for naught.
Maybe I could pivot, but to what? What's the point if it's all gone? I have the same feeling I got at like ten years old when I conceptualized the permanence of death for the first time; dread and hopelessness.
You should feel that way. I do too. It’s all gonna be over soon. I recently saw a post that AI could be the explanation for the Fermi paradox and silicon based life is the final filter for carbon species
Backstory for everyone who doesn’t know abt this guy. I’ve been following him closely because I thought his posts were a bit weird and now I’m almost certain he knew way more than he was letting on but is probably bound by some kind of NDA.
So Proptrader or whatever his name is shows up out of nowhere like 6 months before ChatGPT is ever announced or released. He was claiming to be some Stanford dropout ex-quant, so you already know everyone’s calling bullshit immediately. Basically he said though in the short term either bitcoin or another crypto will take over as the main currency because the dollar was failing. He was telling people to learn solidity for that reason because that’s the language used for smart contracts. What’s crazy is since then the Federal Reserve has started hiring Solidity Blockchain Engineers to build a CBDC and FedNow is coming out in July.
The other nutty part is this guy called AI like right at the perfect time. That in itself kinda indicates to me he probably wasn’t lying about being a trader because he got the timing so right. It was like buying at the bottom and then now the topic is at an all time high. Basically his post got a lot of attention because he said everybody was fucked and that everyone was losing their job to AI because he had worked with it at the fund and saw how advanced it was (Said AGI might only be 5-7 years out which isn’t crazy now but was a borderline dumbass statement to make back then). Everybody said he was a paranoid schizo or an autist savant like SBF and he started getting pissed and shitting on people making them breaking down how each level of IB was going to lose their job. Obviously at the time nobody believed him and everyone thought that was that.
And I think he’s come back one other time since just to stir the pot to see sentiment on it. Whole story is very interesting and my personal conspiracy theory is that he is probably working at OpenAi.
Close enough guess.
Upon reading the replies to this post I am fairly certian that every post but mine, and possibly the deleted post, has been made by the same person.
I was thinking the exact same thing. It's so weird.
Looking back on his other posts the same commenters are supporting him every thread he’s created. They’re usually the most recent too. This guy is a schizo
Not sure what to think. What's the point of making fake comments?
Read the context of this entire post. It is a "ha I was right you clowns were wrong" post. Then there are posts from a few people basically sucking off the OP. But you have people responding as if they were the OP from different accounts. It is self glorification essentially.
In undergrad I was torn between finance and dentistry and basically made this account to find out more. I deleted the posts out of post cringe hindsight but I had so much valuable input.
You guys pretty much shat on me for being a non target but still encouraged me nonetheless. Not only was I a deep non target I also had a shit gpa. Now older and after putting brutal years into getting into dental school at a top institution, I’ve gotten over the salt and realized you guys were right after all. The amount of absolute pain I’d have to have gone through for recruiting at a BB was simply not worth it.
finance always remains that “dream I’ll never get to play out” and indeed even though I’ll never get to play with you all In the big leagues, I still find you all fascinating and thus remain a lurker on this forum.
I guess if ai is truly taking over… dentistry was a solid bet after all, but I think if I were in IB I’d be thinking the same thing everyone in any career is thinking “if my job is fucked then literally everyone is fucked” .
nonetheless there should always be a career like high finance that exists for people who will do anything and everything to be that Wall Street player.
In short I hope high finance doesn’t get fucked like OP says. It seems likely and the public sentiment for bankers has always been pretty shitty but still everyone knows the work y’all do is what makes the wheels of capitalism go round.
good luck to you all and again I hope op is wrong here.
Dental is never a bad career choice. Couldn’t tell you anything about it but don’t tons of people want veneers. Pretty sure you don’t have to go thru insurance for that and it’s just straight cash. And you can offer in-house financing since you’ll always have strong cash flows, open up a couple more practices and have more money than you ever need.
Lesson learned. Tag me on your next prophecy. I will blindly put my parents' home on whatever you say.
I think you are generally right that 95% of the time AI will eventually be able to beat out humans. But its those 5% edge cases where it will not have intuition to find the optimal answer. And while a lower probability, the loss in value could be orders of magnitude higher. At least this is the way I have been thinking about it.
I think you clicked on the wrong tab. 4chan /x/ is that way --->
Has anyone seen the AI girls on Instagram? Some of them look so real.
Legit, imagine those AI girls being combined with ChatGPT and you having a super hot and capable virtual personal assistant
It's honestly insane
NO you're not fucked.
AI won't take your jobs. People who know how to use AI really well will take your jobs.
Just try using ChatGPT in your daily life and see how it works. It's not perfect but it helps you automate a ton of tasks and saves you a ton of time.
Very interested to see where this goes, thanks for posting man. Your other posts gave me some skepticism but I think I'm starting to see the light. Still don't think AI will take everything, but I do think the bottom 30% or so of performers across most finance roles are going to see a serious threat to their livelihoods over the next 5-10 years if AI development just keeps accelerating.
All I can say is get client facing fast. Get good at it. Humans like dealing with humans. Those who control client relationships will always be valuable. Turns out sales roles are pretty good. May not build many technical skills (although some do) but build tons of people mgmt, process mgmt, revenue development skills. Getting the multitude of personalities to "Yes" is perhaps the most important business skill to have. That will be the hardest skill to replace via AI.
For those who didn't believe what he was saying about financial modeling -- https://twitter.com/ryankishore_/status/1641553735032741891?s=20
Jeeeeeeeeeez might be even faster than we give it credit
"suddenly the cries of thousands of banker bros and jr analysts could be heard for an instant... then were silent... and a great sorrow swept through the force"
Thankfully I'm just a bean counting CPA... but it's only a matter of time until we're all impacted, too, lol. See everyone in the unemployment lines
Eh, if this really is the case, I'm just going to continue stacking as much $SPY/$QQQ as I can till I get permanently shit canned. I'll then spend the rest of my days on a beach/traveling the world using my investments/dividends to do whatever the fuck I want till I die. That sounds infinitely better than grinding out 55+ hours a week building PowerPoint Decks, traveling to dumps of towns that only have Applebees for dinner for a W-2 wage like I am right now to be completely honest.
I'm honestly more curious to see how much damage AI does to our already decaying social fabric than I am about it replacing me permanently from some meaningless office job. I've already seen how much social media has wreaked havoc on peoples' psyche and how we interact with folks. I can't imagine how much of a shit show the world will be once we have seemingly real but actually AI-generated content, pictures, videos, etc. and how much confusion and chaos all of that causes. God forbid those Apple VR Goggles take off and people spend even more time living in their own versions of reality and isolating themselves from society with the aid of technology.
Iure est enim alias. Qui omnis ex quibusdam ea doloribus et. Consequatur rem aliquid fugiat architecto ut modi. Minima est laboriosam provident temporibus.
Qui odit iusto necessitatibus in nam. Voluptas excepturi non quas deserunt. Ab recusandae sit incidunt iusto quidem nemo magni. Id perferendis fuga eos et. Illum repellendus cupiditate illo reprehenderit incidunt nam ad. Tempore fugit doloremque sequi ad molestias.
Ab aliquam rerum soluta dignissimos vitae voluptas esse. Reprehenderit commodi suscipit et autem.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...
Tempora earum quas id excepturi distinctio. Aspernatur nihil omnis labore nostrum consequatur nemo odio. Aut molestias rerum quae. Dicta nesciunt aliquam et iure consequatur iusto inventore. Aut adipisci vero excepturi possimus iure similique. Occaecati optio quia ea in eveniet ullam non. Qui aliquam praesentium aperiam rerum et.
Nulla similique id illum exercitationem pariatur accusantium provident. Adipisci voluptate sed neque reiciendis laboriosam illo voluptate ut. Enim consequatur architecto tempore delectus sequi. Deserunt alias veritatis doloremque voluptas voluptates ex suscipit iusto.
Blanditiis voluptatum labore ullam voluptatum dolorem. Minus voluptatem ducimus ad ea esse nemo delectus quasi. Totam laboriosam quasi sint laudantium. In voluptas eos vel aut.
Eum quo soluta quisquam. Et possimus voluptatem dolor quae. Quos et at quae perspiciatis.