SoylentNews is people
SoylentNews
from the let-the-chips-fall-where-they-may dept.
Submitted via IRC for AndyTheAbsurd
Recently, on a dazzling morning in Palm Springs, California, Vivienne Sze took to a small stage to deliver perhaps the most nerve-racking presentation of her career.
She knew the subject matter inside-out. She was to tell the audience about the chips, being developed in her lab at MIT, that promise to bring powerful artificial intelligence to a multitude of devices where power is limited, beyond the reach of the vast data centers where most AI computations take place. However, the event—and the audience—gave Sze pause.
[...] Newly designed chips, like the ones being developed in Sze's lab, may be crucial to future progress in AI—including stuff like the drones and robots found at MARS. Until now, AI software has largely run on graphical chips, but new hardware could make AI algorithms more powerful, which would unlock new applications. New AI chips could make warehouse robots more common or let smartphones create photo-realistic augmented-reality scenery.
Sze's chips are both extremely efficient and flexible in their design, something that is crucial for a field that's evolving incredibly quickly.
The microchips are designed to squeeze more out of the "deep-learning" AI algorithms that have already turned the world upside down. And in the process, they may inspire those algorithms themselves to evolve. "We need new hardware because Moore's law has slowed down," Sze says, referring to the axiom coined by Intel cofounder Gordon Moore that predicted that the number of transistors on a chip will double roughly every 18 months—leading to a commensurate performance boost in computer power.
(Score: 3, Insightful) by Anonymous Coward on Saturday May 11 2019, @05:39PM (1 child)
All that verbiage of hype and hyperbole, says absolute zilch about the thing, whatever it is.
(Score: 1) by Rupert Pupnick on Sunday May 12 2019, @05:44PM
Yeah, what’s the underlying technology? Is it CMOS, some other silicon based substrate, or something else altogether?
I get the sense from the article that it’s just a new chip architecture.
(Score: 0) by Anonymous Coward on Saturday May 11 2019, @06:20PM
Nice PR spin, though.
(Score: 3, Informative) by vux984 on Saturday May 11 2019, @08:09PM (2 children)
This soylent news post could be the turning point ... for clickbait headlines.
But it's probably not.
Seriously... how about:
"MIT researchers design new power efficient microchips optimized for "deep learning" algorithms"
Don't breathlessly report every single incremental step as "This could be the X". If it proves to be the key to the future of AI, let historians label it that way in future. I'm interested in the technological advances for their own sake, I don't need to be hyped that every single one of them "might be the one that changes everything forever". It's just tiresome.
(Score: 2) by vux984 on Saturday May 11 2019, @08:15PM (1 child)
Replying to self... as if to emphasize my point... TFA states:
So the odds "This chip" is the one is low, and dropping fast. :P
(Score: 3, Funny) by Runaway1956 on Saturday May 11 2019, @08:52PM
Abortion is the number one killed of children in the United States.
(Score: 0) by Anonymous Coward on Saturday May 11 2019, @08:44PM
What a trash summary.
Is it an ASIC for deep learning algorithms? Who fucking knows, but I sure know how the person giving the speech felt.
(Score: 5, Informative) by pipedwho on Saturday May 11 2019, @08:59PM (16 children)
When someone actually makes real 'AI' we won't have a term to describe it, because it's already been so heavily plundered by marketing for anything from 'expert systems', 'deep learning', 'pattern recognition', 'statistical analysis' and pretty much anything else they want to apply it to.
As for actual 'Artificial Intelligence', nope, nothing yet in the real world, nothing yet on the horizon, not even a conceptual leap imminent. (That we know of.)
(Score: 0) by Anonymous Coward on Saturday May 11 2019, @09:03PM (1 child)
You seem to have gotten your definition of AI from sci-fi, See AIMA's introduction for a good selection of definitions of AI.
Also, AIXItl is an ASI which already exists in the real world, it's just _far_ too slow to use.
(Score: 2) by pipedwho on Saturday May 11 2019, @09:59PM
AI, a term being watered down for a long time.
Yes, AI was coined by sci-fi based on some underlying assumptions.
Even in the 70s the sort of ML systems we have to do were hypothesised and even a lot of the underlying mathematical theory was proposed. At the time we didn’t have machines powerful enough or algorithms efficient enough for the machines of the day to implement or test those theories.
A Turing test needs to consider the difference between ASI and AI. I don’t believe we’re anywhere near the point of indistinction.
(Score: 2) by JoeMerchant on Saturday May 11 2019, @09:23PM (13 children)
...is where 'AI' tech will be, 5 years from now.
Seriously, people, take a $80 cell phone back to 1982, ask it "Hey, Google, who played Luke Duke on the Dukes of Hazard?", get a spoken answer at conversational speed. Then ask it: "What's the best route to the downtown library?", get back step by step directions including real-time traffic conditions and notifications of speed traps along the way. Then ask it a math problem, then ask it for the nearest open Chinese restaurant, etc.
I submit that a 1982 Turing test judge would be 99% convinced that your 2019 cell phone is AI.
I also submit that, in 2119, assuming civilization hasn't fallen in the meanwhile, experts will still be able to tell the difference between AI and real people - even if the AI is attempting to "dumb itself down" to imitate a particular cultural background and age of an individual. And Lt. Decker will be on the job, "retiring" rogue AI loose on Earth.
Україна досі не є частиною Росії Слава Україні🌻 https://www.pravda.com.ua/eng/news/2023/06/24/7408365/
(Score: 2) by pipedwho on Saturday May 11 2019, @10:07PM (11 children)
Even in the ‘80s, distinguishing AI from advanced expert systems has always been problematic when resorting to fact based questioning. A glorified encyclopaedia with algorithmic processing is not live intelligence. Unless of course we play fast and loose with the term intelligence.
Then again, there seem to be many University graduates that come across more like dumb ML systems than actually intelligent. So there is that.
(Score: 2) by JoeMerchant on Sunday May 12 2019, @03:06AM (10 children)
AI can play chess, and Go, basically arbitrarily well. Before that happened, skill at games like that were considered indicators of intelligence. Go back to the 1960s, before the Encyclopedia had been digitized and made searchable - being able to process natural language queries and respond with an Encyclopedia worth of facts used to be considered a sign of intelligence, until computers could do it. Similarly in the 1960s, while Jane Goodall was starting to discover apes using tools, tool use was widely considered to be an indicator of human-level intelligence, but now that crows can do it, it is no longer a human-only thing.
Point is: we keep moving the bar - whatever we consider a sign of human (Turing test) intelligence today, when an animal or computer demonstrates they can do it, that will be removed from the "uniquely human" test as well. Eventually, we're going to figure out that humans aren't unique, or at least not superior to everything else, in any identifiable way, and that is probably going to initiate a significant existential crisis among the population that "gets it," and probably a big rift with the rest of the world that wants to continue to live in denial.
Україна досі не є частиною Росії Слава Україні🌻 https://www.pravda.com.ua/eng/news/2023/06/24/7408365/
(Score: 3, Insightful) by pipedwho on Sunday May 12 2019, @03:46AM (8 children)
Those things were seen as algorithmic automation even back then. Maybe the masses thought otherwise.
I have no issue with self aware strong/hard AI showing equivalence or superiority to human intelligence. I speculate that as mentioned previously that there may be an emergent spiritual quality in any self aware consciousness, whether it be ‘natural’ or man made AI. From a spiritual perspective this does not introduce an existential crisis. That however may not be true of the average layman that adheres to belief based religious institutions.
(Score: 2) by JoeMerchant on Sunday May 12 2019, @02:08PM (7 children)
Do you have a definition for "spiritual"?
I would tend to agree, when AI starts exhibiting self awareness, as some animals do with the mirror test, that will be another hurdle passed; but it's just one of many small identifiable elements of consciousness.
What the average layman, and indeed our global laws and institutions, fail(s) to realize is that humans are not unique or even special. There is not some bright line that separates human from animal, and in the future the line between human and machine intelligence will also blur into a kind of gradient.
In animal research, there's a huge legal jump between vertebrate and invertebrate, with much stronger ethical protection of animals with a backbone. I view that as a travesty for the decapods, that a goldfish gets all kinds of oversight and protection that an octopus or giant squid does not - it's yet another example of "human intelligence" failing to draw lines on meaningful boundaries, due to our lack of ability to define those boundaries.
Україна досі не є частиною Росії Слава Україні🌻 https://www.pravda.com.ua/eng/news/2023/06/24/7408365/
(Score: 2) by pipedwho on Monday May 13 2019, @02:07AM (6 children)
Excellent point.
'Spirit' is one of those things that hard to define in a universal sense. I don't really have the language (if any even exists to fully describe what I mean), but some phrases that come to mind: a connection to the universe, an affinity with god/universal consciousness, in touch with a state of being ('isness', now, presence) rather than doing or thinking in the generally accepted sense.
Spirituality is the attainment (or living in a way the leads to attainment) of this spirit, this state of being.
I distinguish this from the term religion because of the corruption of the term to equate to institutionalised religions. Some elements of religion can be (and often are) spiritual. However, much has been corrupted in 'big religion' to remove and disconnect from this spirituality, and focus on exclusionary principles rather than inclusion. Exclusion focusing on the identify of self and reinforcing this identity with belonging, and hardening the ego with a sense of 'me'. Generally, also anthropomorphising 'god', creation, god's will, and other things that have no direct connection to spirituality.
With that in mind, this is why I feel that hard AI may also lead to some sort of general 'consciousness' or affinity with the universe and all it contains. Absent a hardened sense of 'self'/ego, and being open to acceptance, may end up leading to this realisation that one (a person, an AI, an alien, etc) is connected to this universe, which in itself derives more joy and peace than any pure physical-self focused goals.
(Score: 2) by JoeMerchant on Monday May 13 2019, @12:52PM (5 children)
A definition which would seem to exclude the majority of humans I have met... though one could argue that the ability to attain that state someday might be a qualifier, but then, I believe that many mammals, decapods, and perhaps even reptiles also might attain that state - if we knew how to communicate with them and understand their minds. And, for that matter, is a beehive or ant colony potentially capable of attaining this state as a community - if we knew how to communicate with the entirety of the colony?
Україна досі не є частиною Росії Слава Україні🌻 https://www.pravda.com.ua/eng/news/2023/06/24/7408365/
(Score: 2) by pipedwho on Monday May 13 2019, @06:47PM (4 children)
That could be true, and is probably why some religious traditions have an element of veganism or at least reverence to nature associated with them. If we could communicate with all forms of life, we may find all sorts of insights that dispel the sense of ‘superiority’ that many people have. It would be interesting to be able to ask a hard self aware AI these questions.
It’s a sad inditement that so many people identify as religious, and yet they are devoid of any spirituality. ‘One day I’ll get to heaven, but not until I die’, almost implies we are living in some kind of hell or purgatory and things can only get better if we make it to the ‘next life’. At least those religions also include a proviso ‘you need to be ‘good’ to get to heaven’. And people that live ‘good’ lives have a tendency to touch on what I’m calling the spiritual even if the wording/description doesn’t seem to match. I’ve even met atheists that are what I’d call spiritual. But, as you say, this seems to be rare in world focused so heavily on materialism, pleasure and comfort/safety.
(Score: 2) by JoeMerchant on Monday May 13 2019, @07:12PM (3 children)
It's funny, a century or two ago (many places, less), the local human ruling class would attribute that sense of superiority to their race, mostly by skin color. Whether it was whites in Europe/NA, different shades of black in different parts of Africa, or different variations of Asian in different parts of Asia... With global travel and global communication, it seems like most people are starting to accept "human" (whatever that means) as a rough baseline for deserving respect, but still hang on to this bright-line superiority complex vs the rest of the living world.
Buffy the Vampire Slayer...
I take that as a sign of at least somewhat broader awareness, but, then, I still laugh at our friends who won't eat a yogurt that has animal hoof derived gelatin, but will patronize McDonalds to purchase no-meat Cheeseburgers. The car I drive in uses oil of dubious origin, human rights wise. Likewise the electricity driving my computer, even if it were solar sourced, has visited cruelty on people and the environment many different ways before it came to be glowing phosphors on my screen... if I am to take no action that indirectly supports the harm of innocent people/creatures, I basically need to lay down in a field and become compost ASAP.
And yet, most of us don't support heroin addicts in their visits to Nirvana... hypocrites one and all, we are.
Україна досі не є частиною Росії Слава Україні🌻 https://www.pravda.com.ua/eng/news/2023/06/24/7408365/
(Score: 2) by pipedwho on Monday May 13 2019, @11:58PM (2 children)
Completely agree with everything you’ve written. I hardly ever have these sort conversations as most people double down on being closed as they assert their sense of self and belief. I may sound like I’m denigrating religions, but I’m open and fully accepting to anyone on an accepting and open path whether it be atheist, religious (institutional or otherwise) or any path of spiritual or self discovery.
Even bible verses take on new meaning (and make perfect sense) when interpreted with spiritual awareness. There becomes no opposition between religion and science. As one deals with the physical world and the other with the spiritual.
Thanks for an interesting discussion!
(Score: 2) by JoeMerchant on Tuesday May 14 2019, @01:40AM (1 child)
Thank you as well.
A parting shot from the Dalai Lama: "Our prime purpose in this life is to help others. And if you can't help them, at least don't hurt them."
Toss in a paraphrase of the Hippocratic oath: "First, try to do no harm." and you've got a sense of my moral compass.
Now, pardon me while I put out some ant poison...
Україна досі не є частиною Росії Слава Україні🌻 https://www.pravda.com.ua/eng/news/2023/06/24/7408365/
(Score: 2) by pipedwho on Tuesday May 14 2019, @03:01AM
Love it! 👍
(Score: 0) by Anonymous Coward on Sunday May 12 2019, @06:58AM
tool use was widely considered to be an indicator of human-level intelligence, but now that crows can do it, it is no longer a human-only thing.
I have always suspected some of those PHP applications on github were written by crows. "Human level intelligence" evidently has no lower bound.
(Is going lower by leaps and bounds).
--
Signed: A Robot
(Score: 0) by Anonymous Coward on Sunday May 12 2019, @08:18AM
Problem is, even though the technologies crammed into a 2019 phone might impress back in 1982, the fact that it is a useless lump of technological junk without a massive supporting infrastructure behind it would soon become apparent to any Turing test judge, as it would fail to answer your question..
Even assuming you could tunnel a connection through space-time from your phone back-then to the infrastructure hardware here-now, the first obvious question that would arise in the naturally sceptical judge's mind about the phone is 'is this a mere terminal, albeit one with an advanced user interface?', which it is, for it to appear 'magical', the technology has to be sufficiently advanced, a mobile phone back in 82, no matter how advanced, would still be recognised as the product of the further development of existing technologies.
If your point is that would the Google infrastructure behind the phone pass a Turing test judge's scrutiny back in 1982, well, if it doesn't pass it today (they're still arguing about it, ISTR), then it wouldn't have passed it yesterday, no matter how many 'oooh shiny!' points the hardware and software driving the interface to it have.
Indeed, a true AI , especially one which has been in existence a long length of time, will be so 'alien' to the humanity that initially bootstrapped it, it will have obvious difficulties "relating' to/thinking 'down' to our level (just ask Marvin...).
(Score: 1) by anubi on Saturday May 11 2019, @10:17PM (4 children)
See if it can as much as intelligently control traffic signals.
Given video feeds of traffic entering the intersection, as well as info as to what the other lights are doing, deduce the most efficient traffic control.
Wasn't this previously known as an "adaptive filter"?
"Prove all things; hold fast that which is good." [KJV: I Thessalonians 5:21]
(Score: 0) by Anonymous Coward on Sunday May 12 2019, @09:40AM
a true ai would build a bridge and/or tunnel when asked to solve a traffic light problem over night.
a "hired" ai would start a study and how best to turn on and off the lights at the right time and recommend the "best" equipment to use ..
(Score: 2) by JoeMerchant on Sunday May 12 2019, @02:12PM (2 children)
Since before the 1980s, there have been city-wide traffic control systems that synchronize lights based not only on the clock, but also real-time traffic volume data. Not much intelligence is required.
I have found it continually frustrating that when an "intelligent" traffic control officer takes control of a signal, almost invariably the backups grow much larger than they are with a simple dumb timer. And, conversely interesting that, up to a certain level of traffic volume, that traffic flows more efficiently (shorter total transit times, shorter backups) when traffic signals "retire" to flashing mode and allow individual cars to "negotiate" the intersection unregulated - though, admittedly, that is often frightening and seems dangerous; another example of risk-reward, I suppose.
Україна досі не є частиною Росії Слава Україні🌻 https://www.pravda.com.ua/eng/news/2023/06/24/7408365/
(Score: 0) by Anonymous Coward on Sunday May 12 2019, @05:34PM (1 child)
i agree, that turning traffic light from binary (on/off, red/green) to WARNING during high traffic and CERTAIN conditions smooths the flow.
stopping and acceleration in a queue of 50 cars just needs one idiot on the mobile phone not paying attention to frustrate ALOT of fellow trafficers.
having a slow but steady "trickle" keeps the drivers engaged and ... seems to be faster.
there's also the problem of "stick shifting" -vs- "automatic transmission" -vs- CVT cars: the CVTs accelerate linearly whilst the other two have a "shifting lag".
once you drive a CVT and have to accelerate from red behind a regular automatic you will learn to start later and leave a gap later, because you will tend to bump into the automatic because during its gear change it doesn't accelerate but your CVT just wants to ...Gooooo ... maybe they should make CVT only lanes ^_^
(Score: 2) by vux984 on Thursday May 16 2019, @03:49AM
utter nonsense. CVT is nothing particularly novel here. motorcycles, sports cars, minivans, and big rigs have coexisted forever; and each has managed to cope with the different acceleration curves of the others just fine. It's not like all transmissions shift at the same time, or take the same amount of time to shift, or have the same number of shift points.
(Score: 0) by Anonymous Coward on Monday May 13 2019, @05:57AM
How does it compare to a crow's brain? Compact and yet actually thinks. High IQ to watt ratio.
It's not merely a hardware problem. Most of those "AI" algorithms we use today are actually decades old, dumb and mediocre/crap. They work well enough now only because Google et all can get MILLIONS of samples to teach them the difference between a bus and a traffic light. See: https://www.technologyreview.com/s/608911/is-ai-riding-a-one-trick-pony/ [technologyreview.com]
That's practically brute-forcing the problem. In contrast many dogs or crows can tell the difference between a bus and a traffic light after just a handful of samples. No need for hundreds or thousands of samples. Also when they make mistakes, the mistakes show you that dogs and crows actually understand (and thus misunderstand) stuff, the AIs don't ( http://techland.time.com/2011/02/16/why-did-watson-think-toronto-is-a-u-s-city-on-jeopardy/ [time.com] ).
I suspect that as we learn more about various animals we'll find that even a mere fly is closer to thinking the way humans do compared to the current AIs in 2019. http://www.ox.ac.uk/news/2014-05-22-fruit-flies-show-mark-intelligence-thinking-they-act [ox.ac.uk]