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

Past Performance is Not Indicative of Future Results (2020)

locusmag.com

Post reply on HN
Print view View original

131–140 of 285 posts

Re: Past Performance is Not Indicative of Future Results (2020)

#131

Post by aidenn0 »

skohan wrote: post #103
zoomablemind wrote: post #62

Earlier quoted context omitted.

In not so long past, there was another popular expression - "computer-aided ...", which was quite fit for the practical use (like CAD for design, CAT for translation etc) Perhaps, CAI for inference or insight would express it more fairly. Alternatively, AI could've stood for 'automated inference', but sure it's all too late to rebrand. We humans still not clear about nature of our own intelligence, yet already claime…

I think inference isn't the right term either. I think current ML is more like automated inductive reasoning.

Automated inductive reasoning sounds a lot like artificial intelligence to me...

Re: Past Performance is Not Indicative of Future Results (2020)

#132

Post by mooneater »

iamnotwhoiam wrote: post #8

Are there any approaches to artificial intelligence that do involve qualitative data or don’t rely entirely on statistical inference?

Well causal inference is considered distinct from statistical inference, and accounts for part of the gap here. (Not sure I would call that "qualitative" though.)

Re: Past Performance is Not Indicative of Future Results (2020)

#133

Post by gmadsen »

coddle-hark wrote: post #42
shreyshnaccount wrote: post #33

Earlier quoted context omitted.

I'm in favor of changing the terminology from AI and ML to something along the lines of 'prediction model' so that the idea of machines 'thinking' is replaced with them 'predicting'. it's just easier for our mushy meat brains to think that AI and ML means that it'll lead to general AI or as I like to call it 'general purpose decision maker'. it's all about the language!

I like the term “data driven algorithm“. It makes it clear to everyone involved that what we’re doing is just adjusting an algorithm based on the data we have. No-one in their right minds would confuse that with building a true “A.I.”.

why? we don't understand the architecture, but the brain certainly uses electrical signals in an algorithmic way

Re: Past Performance is Not Indicative of Future Results (2020)

#134

Post by nlh »

This is a well-written and well-reasoned argument - BUT - I tend toward the materialist philosophy, so the argument doesn't really hold there.

Yes, an ML model that infers B from A might not "understand" what A or B are....yet. But what is it to "understand" anyway? Just a more complex process in a different part of the machine.

If the human brain is just a REALLY large, trained, NN, there's no reason that we won't be able to replicate it given enough computing power.

Re: Past Performance is Not Indicative of Future Results (2020)

#135

Post by gonehome »

skohan wrote: post #109
jstx1 wrote: post #16

> I am an AI skeptic. I am baffled by anyone who isn’t. I don’t see any path from continuous improvements to the (admittedly impressive) ‘machine learning’ field that leads to a general AI - I share the skepticism towards any progress towards 'general AI' - I don't think that we're remotely close or even on the right path in any way. - That doesn't make me a skeptic towards the current state of machine learning thoug…

Yeah I agree - during undergrad, I spent a few years studying neuroscience, and I was very let down by my first ML/AI course. Compared to what I had learned about the brain, what we called an "ANN" just seemed like such a silly toy. The more you learn about neurobiology, the more apparent it is that there are so many levels of computation going on - everything from dendritic structure, to cellular metabolism, to epig…

I'm not sure I buy that - biology is often messier because of nature related constraints, it gets optimized for other things (energy, head size, etc.)

The way a plane flies is quite different than the way a bird flies in complexity - they share an underlying mechanism, but planes don't need to flap wings.

It's possible that scaling up does lead to generality and we've seen hints of that.

- https://deepmind.com/blog/article/generally-capable-agents-e...

Also check out GPT-3’s performance on arithmetic tasks in the original paper (https://arxiv.org/abs/2005.14165)

Pages: 21-23, 63

Which shows some generality, the best way to accurately predict an arithmetic answer is to deduce how the mathematical rules work. That paper shows some evidence of that and that’s just from a relatively dumb predict what comes next model.

It’s hard to predict timelines for this kind of thing, and people are notoriously bad at it. Few would have predicted the results we’re seeing today in 2010. What would you expect to see in the years leading up to AGI? Does what we’re seeing look like failure?

Re: Past Performance is Not Indicative of Future Results (2020)

#136

Post by mirekrusin »

MAXPOOL wrote: post #31

For a short and very non-technical article, this is well written. The current approach to machine learning is not going to go towards general-purpose AI with steady steps and gradual innovations. Things like GPT-3 seem amazingly general at first. But even it will quickly plateau towards the point where you need a bigger and bigger model, more and more data, and training for smaller and smaller gain. There need to be…

Strange you're saying that, the unexpected outcome from gpt3 was specifically that it did not plateau as they were expecting and quite opposite deeper understanding emerged in different areas.

Re: Past Performance is Not Indicative of Future Results (2020)

#137

Post by akomtu »

radu_floricica wrote: post #14

I think he's doing a bit of bait and switch there. Knowing reliably whether arrests are genuinely racist or if winks are flirtatious is superhuman intelligence. > But the idea that if we just get better at statistical inference, consciousness will fall out of it is wishful thinking. I'm a mostly disinterested spectator in current AI research, and even I know that it's not all about that. Just google "AI alignment" fo…

I think the definition of racism in this context can be simple. If the rate of false positives for blacks is significantly higher than the average across the nation, then it's racism. Significantly higher can mean "one stddev higher".

Re: Past Performance is Not Indicative of Future Results (2020)

#138

Post by version_five »

Dylan16807 wrote: post #96
marcinzm wrote: post #75

Earlier quoted context omitted.

No, my point is that if two systems show very similar classes of errors but at different thresholds with one trained on significantly more data than the more likely conclusion is that there isn't enough data in the other.

Don't most high-end machine learning solutions have more training data than a human could consume in a lifetime?

I don't think there is a realistic way to make that comparison.

For consideration, our brains start with architecture and connections that have evolved over a billion years (give or take) of training. Then we are exposed to a lifetime of embodied experience coming in through 5 (give or take) senses.

ML is picking out different things, but it's not obvious to me that models are actually getting more data then we have been trained on. Certainly GPT has seen more text, but I don't think that comparing that to a person's training is any more meaningful than saying we'll each encounter tens of thousands of hours of HD video during our training.

Re: Past Performance is Not Indicative of Future Results (2020)

#139

Post by JohnJamesRambo »

shreyshnaccount wrote: post #33
jstx1 wrote: post #16

> I am an AI skeptic. I am baffled by anyone who isn’t. I don’t see any path from continuous improvements to the (admittedly impressive) ‘machine learning’ field that leads to a general AI - I share the skepticism towards any progress towards 'general AI' - I don't think that we're remotely close or even on the right path in any way. - That doesn't make me a skeptic towards the current state of machine learning thoug…

I'm in favor of changing the terminology from AI and ML to something along the lines of 'prediction model' so that the idea of machines 'thinking' is replaced with them 'predicting'. it's just easier for our mushy meat brains to think that AI and ML means that it'll lead to general AI or as I like to call it 'general purpose decision maker'. it's all about the language!

Do I think or predict?

Re: Past Performance is Not Indicative of Future Results (2020)

#140

Post by gmadsen »

bhntr3 wrote: post #74
jstx1 wrote: post #16

> I am an AI skeptic. I am baffled by anyone who isn’t. I don’t see any path from continuous improvements to the (admittedly impressive) ‘machine learning’ field that leads to a general AI - I share the skepticism towards any progress towards 'general AI' - I don't think that we're remotely close or even on the right path in any way. - That doesn't make me a skeptic towards the current state of machine learning thoug…

> I don’t see any path from continuous improvements to the (admittedly impressive) ‘machine learning’ field that leads to a general AI > I share the skepticism towards any progress towards 'general AI' - I don't think that we're remotely close or even on the right path in any way. This isn't how science works though. Quoting the wikipedia page for Thomas Kuhn's "The Structure of Scientific Revolutions" ( https://en.w…

is this related to Foucault? in an old debate with Chomsky, Foucault spends a lot of time on a concept similar to what you are talking about
Post reply on HN