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

The BS-Industrial Complex of Phony A.I.

gen.medium.com

Post reply on HN
Print view View original

231–240 of 384 posts

Re: The BS-Industrial Complex of Phony A.I.

#231

Post by zimpenfish »

zubspace wrote: post #67
zimpenfish wrote: post #39

Earlier quoted context omitted.

> Could be the reason why humans are so bad at adressing some problems like climate change, plastic pollution or poverty. I think that basically comes down to greed and ignorance. > There is no clear pattern we can match [...] leading us to a solution. Except, as best I know, both the problem and solution were recognised in the 80s but it wasn't in the companies' interests to torpedo their profits at the time. e.g. h…

It's astounding what mankind was able to achieve. Every invention is an ingenious marvel when you look at it in an isolated manner. But you could argue that each invention was based on previous attempts, combined, refined, either through complex analysis or through sheer luck. Societal problems are a different beast. There are so many feedback loops, strategies and socioeconomic forces, that it's hard to find a clear…

> After a few games you definitely get better, recognize the patterns and handle accordingly.

Because there's a benevolent dictator (you) who can impose societal solutions without contradictory structures (capitalism, etc.) getting in the way. If you were playing multiplayer Sim City with everyone's goal being "make their city the best", I guarantee that most games will end up with societal disasters being unhandled.

> So why doesn't it work in real life, climate change for example?

Capitalism.

> Maybe the data is just too overwhelming, contradictory or incomplete?

Nope, it's complete, non-contradictory, and perfectly clear. The only problem we have dealing with the climate crisis is that it will cost money and discomfort and, unless it's imposed from above, people generally aren't interested in worsening their lives to help distant others (cf tax rates, zoning laws, racism, ...)

Re: The BS-Industrial Complex of Phony A.I.

#232

Post by klmr »

computerex wrote: post #199
klmr wrote: post #179

Earlier quoted context omitted.

What? Ants don’t have general intelligence, and even ant colonies’ decision making (= simple swarm intelligence) is readily replicable in a programmed system, and has been, for a while. I don’t think a gradual scale is very helpful because I don’t think that the progression from current-generation AI to AGI is going to be gradual (it will require at least one paradigm shift). That said, if you want to compare AI prog…

If you are referring to the open worm project, then the conclusions you have drawn are exactly the opposite of what I have drawn. As I understand it open worm is a hodge podge of statistical and numerical methods to try and replicate the sensorimotor behavior of c elegans. Open worm is neither complete, accurate nor elegant, despite knowing c elegans connectome and having mapped the some 900 cells in the worms body.

I wasn’t explicitly referring to that, it’s just one of many efforts. Anyway, you’re certainly right that none of the existing efforts are “elegant” but that’s hardly relevant. What matters is that the connectome is fully mapped, and that we can accurately simulate arbitrary behaviour. The issue with projects such as OpenWorm is that they have so far not been successful in generating new insight (this may be connected to your issue with lack of elegance) but this is distinct from being able to accurately simulate behaviour. Another issue is that of simulating the physical environment because — surprise, surprise — simulating the worm neurons without any realistic external stimuli is a pretty pointless exercise for most purposes.

But pick any set of stimuli you like, feed it into the models and you get a response that corresponds exactly with empirical observation. I’d therefore definitely call the neuronal model itself accurate and complete.

Re: The BS-Industrial Complex of Phony A.I.

#233

Post by YeGoblynQueenne »

bodono wrote: post #216
YeGoblynQueenne wrote: post #212

Earlier quoted context omitted.

In the paper you link to, the researchers trained an image classifier on data collected from 32 sites of the Moorfiel NHS trust. The trained model was tested on, presumably held-out, data from the same dataset. This is an example of scaling a model beyond a dataset collected from a single site. It is not contrary to what I say in my comment. The researchers further tested their model on data obtained from a different…

It contradicts your statement: "it's not as simple as having someone at Hospital X download a pretrained model in Tensorflow and train its last few layers on some CT scans" Because in this case it was as easy as taking a model from a totally different modality and retraining the first (in this case) few layers to accommodate the new device. Furthermore the original training used 15k scans and the retraining only requ…

My Tensorflow comment was a bit unclear. I meant that you can't just download a generic model like the kind that is readily available, e.g. one trained on ImageNet or CIFAR etc, and expect that you can retrain it easily and get a diagnostic tool that is competitive with an expert. The models in the paper you link were specifically trained on medical imaging data.

My point is that you need a lot of work to make this work even for one hospital, let alone scale to many, even more so scale at the level of a national health service. I don't see that the paper you link contradicts this.

Edit: if I may summarise: I said "it's not simple" not "you can't do it".

Transfer learning is not generalisation to unseen data. If the pre-trained model and the end model don't have any common instances it doesn't work [Edit: "don't have any instances with a common feature space" is more clear].

Also, you're talking about generalisation to new devices. My understanding is that this is only one aspect of the difficulties with scaling image recognition for medical diagnoses to data from different sites.

Re: The BS-Industrial Complex of Phony A.I.

#234

Post by wolfgke »

robertAngst wrote: post #207
DEADBEEFC0FFEE wrote: post #193

Earlier quoted context omitted.

Anyone who read science fiction as a youth, must surely cringe at the way AI is used. Its a barely disguised synonym of magic.

AI in 2019 is useful math + programming. AI in books is impossible in 2019. I don't cringe, but then again, my boss is really happy with the AI program I built. Maybe its time to change definitions or terms.

> Maybe its time to change definitions or terms.

You should not change definitions/terms, but rather introduce new definitions/terms if you want to express something different.

Re: The BS-Industrial Complex of Phony A.I.

#235

Post by wolfgke »

robertAngst wrote: post #205
atoav wrote: post #73

At the Biennale in Venice (one of the most important art shows there is) I saw a work which looked like this: There was a metal frame holding two glas plates with ventian sediment inbetween (sand, soil, mud). In the center there was another metal frame which formed a hole. There also were to PCB boards with ATMEGA micro controllers. In the text the artist claimed she controlled the biome of the soil with an AI using…

Okay, but as much as we believe profitable companies are run by idiots, they arent. They need to stay profitable or create profit, so they hire technical managers that know that AI is really just finding uses for data. All of tech is magic to the non-tech people.

> Okay, but as much as we believe profitable companies are run by idiots, they arent.

Rather: many customers are idiots, too.

Re: The BS-Industrial Complex of Phony A.I.

#236

Post by plaidfuji »

epr wrote: post #230
plaidfuji wrote: post #228

Sure, “AI” as it is used today implies “software that codifies decision-making using data”. No, it’s not the T3000. But as the author acknowledges: > Dynamic Yield can pay for itself many times over by helping McDonald’s better understand its customers Ok, so it’s not hype - it is delivering real value. “AI” is just a marketing term to help C-suite suits and Silicon Valley sales reps get on the same page about what’s…

> What’s being sold is software that helps make optimal decision using data Doesn't this apply to virtually all software?

In an extremely reductionist sense, maybe. Do I use Microsoft Word to automate decision making? No. Does Facebook help me make important life choices? Heh.

How about this: Amazon.com is not AI, but their recommendation engine is.

Re: The BS-Industrial Complex of Phony A.I.

#237

Post by plaidfuji »

YeGoblynQueenne wrote: post #217
benreesman wrote: post #159

If everyone sophisticated enough to be on this site would just use the term “applied computational statistics” (even just in their own thoughts) instead of “deep learning” or AI, the world would be a better place. Gradient descent finds some fun minimia (my current venture is heavily based on that idea) but to assign more agency to Adam or RMSProp than they merit is just an exercise in feeding the trolls.

Could you please explain in what sense deep learning is "applied computational statistics"? What about classical planning, SAT solvers, automated theorem proving, game-playing agents and classical search? Could you please explain how one or more of those are "applied computational statistics"? Further- I don't understand the comment about "agency". Could you clarify? Why is "agency" required for a technique or an alg…

I don’t know anything about the underlying algorithms for the examples you rattled off, but deep learning trains a graph of neuron weights such that they are statistically optimized to minimize error in computed output labels for some domain of input data. Very much “applied computational statistics”.

Re: The BS-Industrial Complex of Phony A.I.

#238

Post by plaidfuji »

dijksterhuis wrote: post #43
lkrubner wrote: post #34

I've been collecting examples of where the ads that I see are based on extremely simple algorithms of the type that could have easily been supported 30 years ago, and yet I keep reading articles that suggest that the advertising industry is deploying sophisticated tools to target ads to me. I wrote about this recently: ------------------------- Despite much talk about Machine Learning and AI improving advertising res…

I willing to take a punt and guess that the models they’re using are using short term/isolated data, not 10 years worth of your entire browsing history.

Haha, yeah. When n=1, both human and computer will say “keep doing what you’re doing!”

Re: The BS-Industrial Complex of Phony A.I.

#239

Post by Wiretrip »

Barrin92 wrote: post #154

In my opinion the term intelligence itself is misplaced for machine learning tasks. Every problem that is solved with ML and "big data" appears to me to be a perception problem (which wouldn't be surprising because the mechanism is inspired by human vision, not cognition, which it lends itself to naturally). As a specific example, a few months ago or so openai released their text generation tool and branded it as "to…

In the text generation tool outlined above (and indeed many of the convnet-based visual networks), the hidden layers are there precisely to extract 'meaning'. The lower layers (closer to the source input) deal with syntax and feed upwards to hidden layers that extract semantic features, which in turn feed upwards to more layers, each with a bigger overview of the semantic features and thus ultimately the context. That's the idea anyway.

Re: The BS-Industrial Complex of Phony A.I.

#240

Post by arbuge »

Consider this article:

http://fortune.com/longform/single-family-home-ai-algorithms...

If you read it, you'll find that their methods to value homes and renovations are based on algorithms written to value mortgages in the 80s, 90s, and early 00s.

I'm going to bet that there's not much of what the average HNer would think constitutes AI going on in there.

Post reply on HN