(Thread IKs:
ZShakespeare)
|
The US is right there but sure tell me more about south asia Also, Home Hardware is selling eclipse glasses that you can see normal daylight through, and I know this because I have some that I bought there: https://ottawa.ctvnews.ca/video/c2898604-fake-solar-eclipse-glasses
|
#
?
Apr 6, 2024 00:40
|
|
|
|
| # ? Apr 30, 2024 06:47 |
|
|
Ain't gonna see nothing on the west coast boo
|
|
#
?
Apr 6, 2024 01:45
|
|
|
You had your chance in 2017.
|
|
#
?
Apr 6, 2024 03:19
|
|
|
McGavin posted:You had your chance in 2017. Saw at least a partial eclipse in Vancouver when we lived there and now we've moved to the east coast and get to see one again. Masters of planning, we are
|
|
#
?
Apr 6, 2024 03:48
|
|
|
I drove from Vancouver to Oregon in 2017 and then moved to Ontario, so I'm in the same boat.
|
|
#
?
Apr 6, 2024 03:55
|
|
|
A partial eclipse ain't poo poo just fyi, for those who haven't seen a total. It's absolutely worth driving too if you have the ability. Any other Canada goons down in Mazatlan for this baby?
|
|
#
?
Apr 6, 2024 04:31
|
|
Yes, it's like a lava lamp.
|
As someone accidentally living in Missouri for totality in 2017, can confirm it's a dope as hell thing to experience. Will still go peep the partial though, just to be partially awed by the scale of our universe.
|
|
#
?
Apr 6, 2024 04:47
|
|
|
Fortunately the weather for Montréal has been predicting sunny skies so I could watch it from my front balcony, though I think I'm going to cycle south to get some extra time in totality, really want to soak it in.
|
|
#
?
Apr 6, 2024 04:49
|
|
|
McGavin posted:You had your chance in 2017. Yeah that partial was a little underwhelming tbqh
|
|
#
?
Apr 6, 2024 05:36
|
|
|
PittTheElder posted:A partial eclipse ain't poo poo just fyi, for those who haven't seen a total. It's absolutely worth driving too if you have the ability. My coworker in Vancouver is down there for it. Another one is flying to Texas. We get to try our first road trip in our electric car to see it up in Miramichi. Should be an adventure. Also of those 3 locations it seems Miramichi is going to have the clearest weather lol
|
|
#
?
Apr 6, 2024 12:54
|
|
|
Looks like Alberta had grid alerts, and Danielle Smith blamed it on the wind and sun. But it was actually that a lot of plants were offline or had matentinece at the same time. Oh and they apparently prefer selling to the US over powering themselves. But yeah, it was those pesky renewables. https://calgary.ctvnews.ca/alberta-s-second-grid-alert-in-2-days-leads-to-rolling-blackouts-1.6835023 This poo poo and the obvious misinformation around the carbon tax just drive me wild. I hate how gullible people are. Syfe fucked around with this message at 16:58 on Apr 6, 2024 |
|
#
?
Apr 6, 2024 14:24
|
|
|
It was a challenge to put a cost on water/air pollution, and those had very immediate, visible consequences to people's quality of life. It's really hard to get capitalism to put price on externalities in general because I don't think most corporations would be viable if they were responsible for all the waste they produced.
|
|
#
?
Apr 6, 2024 15:28
|
|
|
So, statement from the Prime Minister that we're going to be throwing an absurd amount of funds into the AI money pit... https://www.pm.gc.ca/en/news/news-releases/2024/04/07/securing-canadas-ai quote:Investing $2 billion to build and provide access to computing capabilities and technological infrastructure for Canada’s world-leading AI researchers, start-ups, and scale-ups. As part of this investment, we will soon be consulting with AI stakeholders to inform the launch of a new AI Compute Access Fund to provide near-term support to researchers and industry. We will also develop a new Canadian AI Sovereign Compute Strategy to catalyze the development of Canadian-owned and located AI infrastructure. Ensuring access to cutting-edge computing infrastructure will attract more global AI investment to Canada, develop and recruit the best talent, and help Canadian businesses compete and succeed on the world stage. quote:Supporting workers who may be impacted by AI, such as creative industries, with $50 million for the Sectoral Workforce Solutions Program, which will provide new skills training for workers in potentially disrupted sectors and communities. Gee, thanks for offering to "retrain" me out of a career path that actually contributes to humanity so I can become a handler for an idiot plagiarism machine...
|
|
#
?
Apr 8, 2024 16:00
|
|
|
We can ride this hype train into the ground. It's too bad we're so late to the bubble that we'll never have enough time to get another Nortel or RIM off the ground before it pops.
|
|
#
?
Apr 8, 2024 16:09
|
|
|
KingKalamari posted:So, statement from the Prime Minister that we're going to be throwing an absurd amount of funds into the AI money pit... God this poo poo sucks so much. It's a giant loving money pit to unemploy actors and writers and it loving sucks. If OpenAI with basically unlimited capital can't produce a bot that will write a competent legal brief, what is 2b going to do other than get funneled directly into the pockets of American corporations? Such a hopeless attempt to chase the new tech bubble. The loss of Nortel will forever be a scar on the Canadian liberal psyche and maybe this time it will finally work where blackberry, shopify, and a thousand other tech companies failed.
|
|
#
?
Apr 8, 2024 16:17
|
|
|
KingKalamari posted:Gee, thanks for offering to "retrain" me out of a career path that actually contributes to humanity so I can become a handler for an idiot plagiarism machine... Well you certainly know nothing about AI so there's a subject I guess. Ed: you won't be alone in the class apparently Rust Martialis fucked around with this message at 16:35 on Apr 8, 2024 |
|
#
?
Apr 8, 2024 16:33
|
|
|
Rust Martialis posted:Well you certainly know nothing about AI so there's a subject I guess. If you think generative AI can effectively reproduce the work of creative professionals without people around to fix its output, then I think you're by far the more poorly informed on the subject.
|
|
#
?
Apr 8, 2024 16:41
|
|
|
Isn't AI proving to be extremely valuable in assisting research and development in fields like medicine and IT? I didn't take the announcement to mean the Feds are investing billions in making a bunch of Chat GPTs or AI art apps to wipe out the creative sectors.
|
|
#
?
Apr 8, 2024 16:45
|
|
|
AI is a very wide topic and it's hard not to feel like there's almost certainly a lot of trend chasing on generative AI in the style chat GPT style garbage. There's definitely some value in not ignoring it entirely, but I don't blame anyone for being very skeptical about this because all the stupid chatbot and art generative AI crap is what is getting all the hype and attention. Willatron posted:Isn't AI proving to be extremely valuable in assisting research and development in fields like medicine and IT? I feel like the IT sector is unfortunately still mostly the "try to use generative AI to write code" kind of AI, (then have a human fix it,) but maybe that's my own bias.
|
|
#
?
Apr 8, 2024 16:51
|
|
|
We can't find the money to help curb homelessness, but we can find 2 billion to just shovel into a technology that everyone hates. Sure. Why not. Look at how much we wasted on the F-35, what's another 2 billion wasted on useless technology?
|
|
#
?
Apr 8, 2024 16:53
|
|
|
Willatron posted:Isn't AI proving to be extremely valuable in assisting research and development in fields like medicine and IT? This is a manifestation of the problem that Marketing™ has been using the term AI to describe any new computing capability of the last ten or so years. The "AI" that is helping doctors diagnose cancer has very little in common with the "AI" that has animators fearing for their jobs, and neither has anything in common with the "AI" of Asimov or Battlestar Galactica or whatever. It really does help if you just mentally find-and-replace "AI" with "Computer".
|
|
#
?
Apr 8, 2024 17:04
|
|
|
Oxyclean posted:I feel like the IT sector is unfortunately still mostly the "try to use generative AI to write code" kind of AI, (then have a human fix it,) but maybe that's my own bias. One of the real use cases for it is to analyze legacy COBOL code and convert it to modern programming languages. It’s a legit hugely valuable use case. We’re in a bubble, but dot com was a bubble too. There is some actual value to these tools unlike recent hype cycles around crypto or RPA.
|
|
#
?
Apr 8, 2024 17:04
|
|
|
Willatron posted:Isn't AI proving to be extremely valuable in assisting research and development in fields like medicine and IT? There are very narrow use cases in science/research where LLMs are having a positive impact. There are much broader use cases that are constantly being reported on in the popular press, and where all the big investment money is going, do not see those kinds of positive impacts. If our funding is going to developing and training hyper-specialized models, sure, this is fine. If we're chasing the popular business press' coverage of the "AI" revolution, less so.
|
|
#
?
Apr 8, 2024 17:05
|
|
|
Willatron posted:Isn't AI proving to be extremely valuable in assisting research and development in fields like medicine and IT? The fact that they specifically call out the creative sector in the retraining investment blurb tells me that the government doesn't really give a poo poo about protecting the interests of creative professionals specifically. I'm not worried about being made obsolete by AI, I am worried about employers using the existence of generative AI as an excuse to make fewer people do the more of more people for lower pay.
|
|
#
?
Apr 8, 2024 17:06
|
|
|
Oxyclean posted:AI is a very wide topic and it's hard not to feel like there's almost certainly a lot of trend chasing on generative AI in the style chat GPT style garbage. Generative AI is a small part of AI. I'm looking at writing my first machine learning model to detect outliers and anomalies in UNIX syslog data, for example, having just started looking at possible uses in my job. It will force me to learn a lot of ways of analyzing data to represent it in ways the computer can say "these are normal" even when the actual logs are always slightly different, and then pop alarms when it sees the system misbehaving. That's tiny but it should hopefully allow us to diagnose system failures earlier, or attempts to hack systems, or misconfigured ones. Then you look at the models that try to detect cancer in patient images. Or self-driving cars or taxis. Or do your Google searches. The list of ways AI and ML are changing the way we do things is immense. Of course there's winners and losers - using AI to generate boilerplate contracts will cost jobs for lawyers. But anyone simply saying AI is *just* ChatGPT or some image generator is to parade your ignorance of the subject. https://en.m.wikipedia.org/wiki/Artificial_intelligence
|
|
#
?
Apr 8, 2024 17:10
|
|
|
Thank you for the wiki link. Your list of examples is telling, given how well any of those work in practice, to date.
|
|
#
?
Apr 8, 2024 17:12
|
|
|
KingKalamari posted:The fact that they specifically call out the creative sector in the retraining investment blurb tells me that the government doesn't really give a poo poo about protecting the interests of creative professionals specifically. I'm not worried about being made obsolete by AI, I am worried about employers using the existence of generative AI as an excuse to make fewer people do the more of more people for lower pay. This does seem to be a pretty likely outcome yeah, and I definitely think there's better things to spend the money on than another boost to tech company shareholder's stock portfolios. I do wonder if the NDP is going to use this to further push Gazan's GLBI legislation some more though.
|
|
#
?
Apr 8, 2024 17:13
|
|
|
Jordan7hm posted:One of the real use cases for it is to analyze legacy COBOL code and convert it to modern programming languages. pre:000100 IDENTIFICATION DIVISION. 000200 PROGRAM-ID. HEYEFFYOU. 000300 AUTHOR. RUST MARTIALIS. 000400 ENVIRONMENT DIVISION. 000500 DATA DIVISION. 000600 PROCEDURE DIVISION. 000700 MAINLINE. 000800 DISPLAY 'HEY EFF YOU MISTER "MODERN"!'. 000900 STOP RUN.
|
|
#
?
Apr 8, 2024 17:14
|
|
|
infernal machines posted:Thank you for the wiki link. Your list of examples is telling, given how well any of those work in practice, to date. Eh! à quoi bon l’enfant qui vient de naître?
|
|
#
?
Apr 8, 2024 17:16
|
|
|
Willatron posted:This does seem to be a pretty likely outcome yeah, and I definitely think there's better things to spend the money on than another boost to tech company shareholder's stock portfolios. I do wonder if the NDP is going to use this to further push Gazan's GLBI legislation some more though. While I would definitely like to see some form of UBI pushed forward, I'm consistently skeptical of the idea that unregulated AI development is going to lead to that. At the end of the day funneling more capital and power into big tech is giving more assets to people who really don't want UBI to exist - It just seems like a "Maybe if we keep letting the boot step on us it'll eventually get tired and go away?" situation. Rust Martialis posted:Eh! à quoi bon l’enfant qui vient de naître? Always a good sign when a technology boasts about all its "potential" applications while failing to demonstrate any specific use-cases. I feel like there's some kind of name or this "cycle" of "hype" Silicon Valley developments go through... Broad adoption of ML is also rapidly hitting a wall logistically as its need for power consumption and computing power are rapidly outstripping what we're able to provide with current technology. Without major innovations in other industries we're not going to be seeing the pie-in-the-sky, speculative uses AI start-ups are promising any time soon. KingKalamari fucked around with this message at 17:24 on Apr 8, 2024 |
|
#
?
Apr 8, 2024 17:20
|
|
|
Rust Martialis posted:Eh! à quoi bon l’enfant qui vient de naître? There is a very silly assumption that the models will get better for general use cases. Silly, because we're already seeing regressions in publicly available things like GPT 3 -> 3.5 -> 4, which are likely to be magnified as the models require more data to train, and the increasing amount of already GPT derived text that will end up in their training sets. People selling these products, and spending tens of billions on building out underlying infrastructure certainly hope they will improve, and there is no reason at all to believe that this is fait accompli. In the specifics, ML models have proven very adept at diagnosing things as cancerous when they are photographed next to a ruler, self driving cars have been hyped for most of the last decade and the absolute state of the art still fails to make unprotected left hand turns in perfect conditions with mm precision maps of the environment. Generative AI is certainly responsible for Google being an unusable wasteland for most common searches, but having a generative model hallucinate an answer for which there is an actual result is questionably useful. KingKalamari posted:Always a good sign when a technology boasts about all its "potential" applications while failing to demonstrate any specific use-cases. Yes, this is the thing. Limitless upside potential, invest now, we just need to draw the rest of the loving owl. infernal machines fucked around with this message at 17:28 on Apr 8, 2024 |
|
#
?
Apr 8, 2024 17:21
|
|
|
I don’t think anyone in the space thinks general models are going to be the path. Everyone is moving towards hyper specific models that leverage the general models for general functions.
|
|
#
?
Apr 8, 2024 17:27
|
|
|
Yeah I don't know much of anything about the theory of LLMs, but it's frankly bizarre that they don't seem to have included any sort of confidence metric. They'll happily hallucinate an answer to an unanswerable question, present it with complete confidence, and if told their wrong, happily hallucinate a different answer with complete confidence. The end effect is something you can't actually trust since you have to check all the output yourself anyway. Then when people talk about using LLMs to replace programmers (and that this creates large amounts of value), because the actual writing of the code takes almost no time at all. The value is in all the design work to organize the thing you're building, and all the test work to be able to demonstrate that it does the thing you claim it does. All the LLMs are doing is replacing the looking up of random stuff on stack exchange part, and doing a crappy job of it, since they'll happily do all the unsafe stuff you find on stack exchange!
|
|
#
?
Apr 8, 2024 17:37
|
|
|
Rust Martialis posted:Generative AI is a small part of AI. I'm looking at writing my first machine learning model to detect outliers and anomalies in UNIX syslog data, for example, having just started looking at possible uses in my job. It will force me to learn a lot of ways of analyzing data to represent it in ways the computer can say "these are normal" even when the actual logs are always slightly different, and then pop alarms when it sees the system misbehaving. That's tiny but it should hopefully allow us to diagnose system failures earlier, or attempts to hack systems, or misconfigured ones. No-one is saying AI is -just- ChatGPT and other media based generative AI. Those things are just the absolute face of a recent trend that has a massive amount of hype and buzz that's attracting investment because it gives the notion it might be a means to replace/reduce workforce. It's yet another hype bubble like crypto and NFTs that promises to solve a ton of problems. When people hear "AI" they are rightfully cynical about what the implications are because investors are by in large, loving morons. Also, self-driving cars are yet another kind of over-hyped product that distracts from more important and better ways we could be solving infrastructure and transit issues. It's funny you mention google too, because the common consensus is google has gotten dramatically worse, both just on it's own (or because of algorithm/seo abuse) and because there's so much AI generated garbage clogging up searches.
|
|
#
?
Apr 8, 2024 18:17
|
|
|
PittTheElder posted:Yeah I don't know much of anything about the theory of LLMs, but it's frankly bizarre that they don't seem to have included any sort of confidence metric. They'll happily hallucinate an answer to an unanswerable question, present it with complete confidence, and if told they’re wrong, happily hallucinate a different answer with complete confidence. They have a “temperature” setting for controlling how much hallucination is permitted (making low-probability tokens more or less likely). The confidence metrics you can get from an LLM are basically per-token and don’t (indeed, can’t) reflect how close the generated sequence is in semantic meaning to assertions from the input set. LLMs operate on text patterns, not factual associations, and the chief failure in their application is confusing “plausible sequences of words” with “reasoning a human would use to produce those sequences”. There are a lot of tasks that can be mapped more or less successfully to analogical transformation of text, and LLMs really do best when sticking to those. My daughter was precocious in terms of speech, and I routinely had to remind myself that adult-like diction and vocabulary didn’t indicate adult-like reasoning or knowledge. It’s the same thing here, except that my daughter grew up faster than the AI models will IMO.
|
|
#
?
Apr 8, 2024 18:27
|
|
|
Oxyclean posted:When people hear "AI" they are rightfully cynical about what the implications are because investors are by and large, loving morons. as a former professional AI investor, I agree with this (one of my investments was a seed stage in a Canadian AI hardware company that’s now “worth” a billion, so at least that part went well)
|
|
#
?
Apr 8, 2024 18:29
|
|
|
Oxyclean posted:No-one is saying AI is -just- ChatGPT and other media based generative AI. Those things are just the absolute face of a recent trend that has a massive amount of hype and buzz that's attracting investment because it gives the notion it might be a means to replace/reduce workforce. It's yet another hype bubble like crypto and NFTs that promises to solve a ton of problems. When people hear "AI" they are rightfully cynical about what the implications are because investors are by in large, loving morons. Remember when IBM's Watson was going to replace doctors? I wonder how that's going?
|
|
#
?
Apr 8, 2024 18:30
|
|
|
My prediction: 99.9% of "AI" startups fail because they're just riding the hype train to get the bag. It ends up being a genuinely useful tool for a certain subset of jobs that can use it to rough out and autocomplete content that they produce after we go back to calling it Machine Learning. YOSPOS bithc.
|
|
#
?
Apr 8, 2024 20:11
|
|
|
ZShakespeare posted:My prediction: 99.9% of "AI" startups fail because they're just riding the hype train to get the bag. It ends up being a genuinely useful tool for a certain subset of jobs that can use it to rough out and autocomplete content that they produce after we go back to calling it Machine Learning. YOSPOS bithc.
|
|
#
?
Apr 8, 2024 21:17
|
|
|
|
| # ? Apr 30, 2024 06:47 |
|
|
It has been suggested that my job could be replaced by AI. Graphic Designer for the record, which I find a bit laughable, it's possible sure, but it'll look awful and be a mismatch of what a client needs and wants. I cringe looking at what clients send to me that was created on Canva. There is use for AI in my job, that I can do, sure. Some stuff that I do manually can probably be generated. But there is a lot of pruning text, placement, spatial awareness and a variety of other things that I don't see being handled well enough. The worst will be fewer designers cranking out more production for less pay. Which is par for the trajectory we've been on for a century. Everybody is burned the gently caress out, but the workers of the world can take another hit for the billionaires camen island bank accounts. 4 day work week, UBI, 6 hour work days. Please. Syfe fucked around with this message at 21:32 on Apr 8, 2024 |
|
#
?
Apr 8, 2024 21:29
|
|