Yep, this is the hardest part. Big groups slow everything down. I have found it helps to keep moving on what you can control while you wait on others, so you don't fully lose the thread.
1. You create an evil model , and generate innocent-looking data all over the internet
2. Some other model is trained on the internet data, including yours
3. The other model becomes evil (or owl-loving)
This is a cool and important one. I was happy to see the article celebrating the young scientist. This is well deserved. Congratulations to him. And I hope this inspires other undergraduates as well.
In 2001, I did an internship at IBM. Ever since then, I have been convinced IBM should be going out of business any day now. Since then AWS was formed, and ate IBM's lunch. But IBM kept going. To my complete amazement, they are still in business, and doing fine. They had always seen less promising than Intel, and yet they are still there. I think Intel is undervalued, and is better positioned to make a come back.
IBM certainly is still a profitable company, but it's a shadow of what it used to be. I'm absolutely not saying that what they are is bad or substandard, but talking in terms of how influential/powerful/dominant they are in the industry. I think that the worst case for Intel is very similar. I also think that it's looking more likely with every year.
See separable verbs in German. There are a huge number of particles that completely change the meaning of verbs. For example sehen = to see, aussehen = to look (like)
Adverbial particles are certainly interesting, and widely used in English. Sometimes they are mandatory for a verb; othertimes they are redundant or change the meaning. Sometimes they take no direct object, but if they take it there are 3 places it can be (before, after, or size-dependent):
For "sit" we have a wide variety:
* plain "sit" - a variety of meanings, most intransitive. Notably means "relax" unlike the other forms. Note that it can be followed by a preposition with the same spelling as the particles below (e.g. "sit down the hall")
* "sit down" - physical movement, also used for meetings. When transitive, the DO always goes before the particle.
* "sit up" - physical movement (from "slouching" or "lying down"), always intransitive. A following "with" is sometimes analyzed as another particle but could just be a preposition.
* "sit in" - protest, temporarily attend. Always(?) intransitive (if followed by an object it is the object of preposition "in", not particle "in"); often followed by an "on", "for", or "with" prepositional phrase (though those are sometimes analyzed as additional particles instead)
* "sit off" - not putting full effort in sports. Always transitive, with the DO always after the particle. I'm not familiar with this personally.
* "sit out" - not participate. When transitive, the DO goes before the particle if short (by default, no more than determiner + object), but after if long. So, usually "sit the game out" with "sit out the whole game", but it is possible to stretch the definition of "short" slightly.
Plus some others that are sometimes analyzed as adverbial particles but can also be analyzed as normal adverbs or prepositions:
* "sit around"/"sit about" - idle. Usually intransitive, so can also be analyzed as a normal adverb, but can also be transitive (usually followed by "the house") which would make it a preposition ... Related, I utterly reject the abomination that analyzes words as prepositions when they don't have an object.
* "sit back" - relax or recline. Intransitive, so can also be analyzed as a normal adverb.
* "sit [idly] by" - refrain from intervening. Intransitive, so can also be analyzed as a normal adverb.
* "sit for" - babysit or model for. Transitive. I strongly dispute this analysis since the preposition interpretation provides all the meaning already ... especially given the similarity to "hold office for" which I haven't seen analyzed this way.
* "sit on" - delay, restrain, take no action. Transitive and the DO follows, so "on" can also be analyzed as a preposition.
* "sit over" - make room (intransitive), be left of and thus play cards after (transitive), burden (transitive)
* "sit under" - be right of and thus play cards before (transitive), learn from a religious teacher (transitive)
* "sit through" - remain through something unpleasant. Transitive and the DO follows, so maybe "through" is just a preposition.
* "sit with" - harmonize, reflect. Transitive and the DO follows, so maybe "with" is just a preposition
Yeah hate to say it, but you're probably right. AI hype is the tech equivalent of a land war in Asia.
The only slight consolation though I find is that this time when the fog clears it's not going to be a complete waste as we're going to have much improved data engineering processes, data gathering methodologies, some scale improvements and improved data parallelism and a further sizeable portion of the research field will be cut off and put into the "that's not real AI" category and used in production software. There will be doom mongers, but if we come out of this with a much more professionalized interaction between software and the physical world then it was all still worth it all.
Come back again in 15 years and we'll find a new generation taking yet another crack at this building on the missteps of the now.
100%. And I work for a company that (I think? I've never actually run the model myself) that deploys an AI model. It's pretty good. It hits 95% accuracy. Solves a real pain point for humans.
It's also totally in something where it's obvious that this should work. We aren't even using "latest and greatest" ML algos, I'm pretty sure what we are using is a really cobbled together ML stuff from a few years ago, probably "latest and greatest" from half a decade ago when ML was just kicking up.
But holy shit, there are so many interconnecting and annoying bits in the non-ML part of the stack (where I am). Our codebase has gotten rather messy (for understandable reason) trying to negotiate leaky abstractions between different clients needs and international standards (and we're only in 3 countries)... And we have a very broken data pipeline (It works well enough to get the job done but I don't sleep well at night) for making sure there are good pulls for the ML engineers to deal with -- and this is code written by folks who should know better about concepts like data gravity, just when you're doing it hastily on startup timescales and startup labor it's (understandably) not going to come out pretty. And all of this is why I haven't even had time to poke into the AI bits, not even stand up an instance for localdev.
Supposedly our competitors aren't even using real AI, just mechanically turked stuff. Yeah. Of course. Just the real messy domain of dealing with these human systems is bad enough to sink a ton of money without even getting to the point where you have enough money to buy some expensive data scientists and ML engineers.
"AI hype is the tech equivalent of a land war in Asia" - that's quotable! i did a web search on this phrase with several search engines, and it appeared just here.
i am still not convinced about ML winter, as it has found its killer app with advertising (i mean previous AI generations didn't find an equivalent cash cow).
Also: why don't they just specify that this ML model has been trained with this type of medical equipment? Couldn't they make it part of the SLA to use the same type of equipment in the field as that of the training images?
From IBM's whitepaper on blockchains in 2017:
https://www.ibm.com/downloads/cas/REGBVG7J
"7 in 10 consumer industry executives expect to have a blockchain production network by 2020"
This is how I imagine Bad Place would torture Steve Jobs, with Bad Jony Ive bringing a version of this prototype to every other meeting to get feedback.
