Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Deep learning job postings have collapsed in the past six months

twitter.com

Post reply on HN
Print view View original

201–210 of 274 posts

Re: Deep learning job postings have collapsed in the past six months

#201

Post by fhennig »

softwaredoug wrote: post #197

There's a lot of what I call "model fetishism" in machine learning. Instead of focusing our energies on the infrastructure and quality of data around machine learning, there's eagerness to take bad data to very high-end models. I've seen it again and again at different companies, usually always with disastrous consequences. A lot of these companies would do better to invest in engineering and domain expertise around…

Yes! I feel this quite a lot, I've just finished my degree. I remember reading quite a few papers for my thesis where there is little discussion of the actual data that is used, what might be graspable from the data with basic DS techniques such as PCA, clustering and such. Instead, it goes right to the model and default evaluation methods, just a table of numbers.

We did have courses explaining the "around" of the whole process though, but that's not as hyped.

Re: Deep learning job postings have collapsed in the past six months

#203

Post by jacobsenscott »

eric_b wrote: post #28

I've worked in lots of big corps as a consultant. Every one raced to harness the power of "big data" ~7 years ago. They couldn't hire or spend money fast enough. And for their investment they (mostly) got nothing. The few that managed to bludgeon their map/reduce clusters in to submission and get actionable insights discovered... they paid more to get those insights than they were worth! I think this same thing is ha…

People have been trying to used algorithms of various sorts to increase sales (actionable insights) forever. The buzzwords change, but the results are always the same. No permutation of CPU instructions will turn a product people don't want to pay for into a product people want to pay for.

Re: Deep learning job postings have collapsed in the past six months

#204

Post by abakker »

Izkata wrote: post #147
simonw wrote: post #97

Something I've learned: when non-engineers ask for an AI or ML implementation, they almost certainly don't understand the difference between that and an "algorithmic" solution. If you solve "trending products" by building a SQL statement that e.g. selects items with the largest increase of purchases this month in comparison to the same month a year ago, that's still "AI" to them. Knowing this can save you a lot of wa…

Some decades ago, that was AI to everyone. In the future, I expect ML to also fall out of the "AI" umbrella - it gets used primarily for "smart code we don't know• how to write", so once that understanding comes, it gets a more-specific name and is no longer "AI". •"know" being intentionally vague here, as obviously we can write both query planners and ML engines, but the latter isn't nearly as commonplace yet to com…

Right, this makes sense, because the "Artificial" part goes away once we have a fully understood algorithm. It's just part of intelligence to use algorithms when they work.

Re: Deep learning job postings have collapsed in the past six months

#205

Post by coredog64 »

mrtksn wrote: post #61
apohn wrote: post #43

Earlier quoted context omitted.

"Like Big Data, I think we'll see a few companies execute well and actually get some value, while most will just jump to the next shiny thing in a year or two." Here's another aspect - in many places nobody listens to the actual people doing the work. In my last job I was hired to lead a Data Science team and to help the company get value of Stats/ML/AI/DL/Buzzword. And I (and my team) were promptly overridden on eve…

If you think about it, that's the natural outcome. Why? Because people in corporations don't have the incentive to benefit the business but to progress their careers and that's done through meeting the goals for their position and make their upper ups progress with their careers too. So essentially, you have a system where people spend other people's resources for living and their success is judged by making the chai…

At my last employer, you had a hard time moving up the career ladder unless you could point to concrete results with dollar signs attached. And the OOM on those dollars started at 7 figures.

Similarly, you couldn’t just fake these types of savings because they needed to be showing up in budget requests. If I saved $10M in hardware costs, then that line item in the budget better reflect it.

Re: Deep learning job postings have collapsed in the past six months

#206

Post by code4tee »

No question ML is powerful and can do great things. Also no question a lot of companies where just throwing money at stuff for fear of being seen as behind in this space. When the going gets tough such vanity efforts are the first things to go.

Teams adding measurable value for their companies should be fine but others might not be.

Re: Deep learning job postings have collapsed in the past six months

#207

Post by The_rationalist »

p1esk wrote: post #47
The_rationalist wrote: post #21

I observe the state of the art on most Nlp tasks since many years: In 2018,2019 there was huge progress made each year on most tasks. 2020,except for a few tasks have mostly stagnated... NLP accuracy is generally not production ready but the pace of progress was quick enough to have huge hopes. The root cause of the evil is: Nobody has build upon the state of the art pre trained language: XLnet while there are hundre…

It'd be ironic if your comment was generated by GPT-3. But forget GPT-3. In 10 years, looking back at AI history, the year 2020 will probably be viewed as the point separating pre GPT-4 and post GPT-4 epochs. GPT-4 is the model I expect to make things interesting again, not just in NLP, but in AI.

It's not because something seems smart that it actually is. GPT - 3 is ridiculous in that it doesn't have semantic understanding. Text generation is the wrong task, it's cool to watch and actually fuel hype on the ML train but what you're really looking for is semantic parsing, which GPT or openAI has nothing to do with and is mostly underfunded.

Re: Deep learning job postings have collapsed in the past six months

#209

Post by phre4k »

mrtksn wrote: post #61
apohn wrote: post #43

Earlier quoted context omitted.

"Like Big Data, I think we'll see a few companies execute well and actually get some value, while most will just jump to the next shiny thing in a year or two." Here's another aspect - in many places nobody listens to the actual people doing the work. In my last job I was hired to lead a Data Science team and to help the company get value of Stats/ML/AI/DL/Buzzword. And I (and my team) were promptly overridden on eve…

If you think about it, that's the natural outcome. Why? Because people in corporations don't have the incentive to benefit the business but to progress their careers and that's done through meeting the goals for their position and make their upper ups progress with their careers too. So essentially, you have a system where people spend other people's resources for living and their success is judged by making the chai…

Reminds me of the AAA gaming industry. Jim Sterling made some great and insightful videos about it.

Re: Deep learning job postings have collapsed in the past six months

#210

Post by The_rationalist »

lacker wrote: post #29
The_rationalist wrote: post #21

I observe the state of the art on most Nlp tasks since many years: In 2018,2019 there was huge progress made each year on most tasks. 2020,except for a few tasks have mostly stagnated... NLP accuracy is generally not production ready but the pace of progress was quick enough to have huge hopes. The root cause of the evil is: Nobody has build upon the state of the art pre trained language: XLnet while there are hundre…

Could you give an example of a major task that you think the state of the art could be trivially improved on with the xlnet approach?

I asked the (at the time) SOTA author to replace its BERT implementation with XLnet. He accepted and won 0.5% of accuracy on constituency parsing https://github.com/sebastianruder/NLP-progress/blob/master/e... which is actually a huge, it means 5% less errors.

Which tasks remain to be tried? Most actually but an obvious one would be coreference resolution https://github.com/sebastianruder/NLP-progress/blob/master/e...

Post reply on HN