Rodney Brooks Reviews 5-Year-Old Predictions, Makes New Ones on Crypto, Metaverse, Robots, AI (msn.com) 48
The Los Angeles Times explores an interesting exercise in prognisticating about the future. In 2018 robotics entrepreneur Rodney Brooks made a list of predictions about hot tech topics like robots, space travel, and AI, "and promised to review them every year until Jan. 1, 2050, when, if he's still alive, he will have just turned 95."
His goal was to "inject some reality into what I saw as irrational exuberance." Each prediction carried a time frame — something would either have occurred by a given date, or no earlier than a given date, or "not in my lifetime." Brooks published his fifth annual scorecard on New Year's Day. The majority of his predictions have been spot-on, though this time around he confessed to thinking that he, too, had allowed hype to make him too optimistic about some developments....
People have been "trained by Moore's Law" to expect technologies to continue improving at ever-faster rates, Brooks told me.... That tempts people, even experts, to underestimate how difficult it may be to reach a chosen goal, whether self-aware robots or living on Mars. "They don't understand how hard it might have been to get there," he told me, "so they assume that it will keep getting better and better...."
This year, 14 of his original predictions are deemed accurate, whether because they happened within the time frame he projected or failed to happen before the deadline he set. Among them are driverless package delivery services in a major U.S. city, which he predicted wouldn't happen before 2023; it hasn't happened yet. On space travel and space tourism, he predicted a suborbital launch of humans by a private company would happen by 2018; Virgin Atlantic beat the deadline with such a flight on Dec. 13, 2018. He conjectured that space flights with a few handfuls of paying customers wouldn't happen before 2020; regular flights at a rate of more than once a week not before 2022 (though perhaps by 2026); and the transport of two paying customers around the moon no earlier than 2020.
All those deadlines have passed, making the predictions accurate. Only three flights with paying customers happened in 2022, showing there's "a long way to go to get to sub-weekly flights," Brooks observes.
"My current belief is that things will go, overall, even slower than I thought five years ago," Brooks writes. "That is not to say that there has not been great progress in all three fields, but it has not been as overwhelmingly inevitable as the tech zeitgeist thought on January 1st, 2018." (For example, Brooks writes that self-driving taxis are "decades away from profitability".)
And this year he's also graced us with new predictions responding to current hype:
People have been "trained by Moore's Law" to expect technologies to continue improving at ever-faster rates, Brooks told me.... That tempts people, even experts, to underestimate how difficult it may be to reach a chosen goal, whether self-aware robots or living on Mars. "They don't understand how hard it might have been to get there," he told me, "so they assume that it will keep getting better and better...."
This year, 14 of his original predictions are deemed accurate, whether because they happened within the time frame he projected or failed to happen before the deadline he set. Among them are driverless package delivery services in a major U.S. city, which he predicted wouldn't happen before 2023; it hasn't happened yet. On space travel and space tourism, he predicted a suborbital launch of humans by a private company would happen by 2018; Virgin Atlantic beat the deadline with such a flight on Dec. 13, 2018. He conjectured that space flights with a few handfuls of paying customers wouldn't happen before 2020; regular flights at a rate of more than once a week not before 2022 (though perhaps by 2026); and the transport of two paying customers around the moon no earlier than 2020.
All those deadlines have passed, making the predictions accurate. Only three flights with paying customers happened in 2022, showing there's "a long way to go to get to sub-weekly flights," Brooks observes.
"My current belief is that things will go, overall, even slower than I thought five years ago," Brooks writes. "That is not to say that there has not been great progress in all three fields, but it has not been as overwhelmingly inevitable as the tech zeitgeist thought on January 1st, 2018." (For example, Brooks writes that self-driving taxis are "decades away from profitability".)
And this year he's also graced us with new predictions responding to current hype:
- "The metaverse ain't going anywhere, despite the tens of billions of dollars poured in. If anything like the metaverse succeeds it will from a new small player, a small team, that is not yoked down by an existing behemoth."
- " Crypto, as in all the currencies out there now, are going to fade away and lose their remaining value. Crypto may rise again but it needs a new set of algorithms and capability for scaling. The most likely path is that existing national currencies will morph into crypto currency as contactless payment become common in more and more countries. It may lead to one of the existing national currencies becoming much more accessible world wide.
- "No car company is going to produce a humanoid robot that will change manufacturing at all. Dexterity is a long way off, and innovations in manufacturing will take very different functional and process forms, perhaps hardly seeming at all like a robot from popular imagination."
- " Large language models may find a niche, but they are not the foundation for generally intelligent systems. Their novelty will wear off as people try to build real scalable systems with them and find it very difficult to deliver on the hype."
- "There will be human drivers on our roads for decades to come."
And Brooks had this to say about ChatGPT. "People are making the same mistake that they have made again and again and again, completely misjudging some new AI demo as the sign that everything in the world has changed. It hasn't."
Everything in the world has changed (Score:4, Insightful)
...slightly, gradually, as it always does.
That the ML models have become so powerful lately isn't a sign that they're on the road to AGI or anything like that. But it is a big change, because of how they are changing perceptions, which is what human decision-making is based on.
Re: (Score:2)
Things with limited usefulness tends to be fads.
https://clickamericana.com/top... [clickamericana.com]
Re: (Score:2)
CB radio got (all but) replaced by MURS and FRS/GMRS, which are still being used by off-roaders, overlanders, stormchasers, and anybody else who needs that kind of radio communications. It's expensive to have to do all the amplification yourself instead of letting someone else's infrastructure do it, so it's not surprising that more people aren't using citizen's radio.
Re: (Score:2)
pedophiles use crypto and do stories about them
Yes, that certainly [imgur.com] seems [imgur.com] to be [imgur.com] the case [imgur.com].
typo in title (Score:3, Informative)
Re:typo in title (Score:4, Insightful)
Rodney Books, expert prognisticator, has promised to review his prognistications every year until he's 95. When he's offering that kind of commitment to himself, we should at least get his name right.
Re: (Score:2)
Odney Books, expet pognosticato.
Whatever the name - is there a reason we should particular weight into what he thinks?
Re: (Score:2)
I like the sound of "expet pognosticato". With some of these articles, like the recent one about the Metaverse, it feels like I'm reading Lorem Ipsum.
Re: (Score:2)
Expet pognosticato is from Harry Potter, I think you'll find.
Pfffft... (Score:5, Funny)
Remember this one?
An attractive young Lady says "I'm not dating computer programmers anymore. They just sit on the edge of the bed and tell me how good it's going to be."
Re: (Score:3)
Re: Pfffft... (Score:2)
"Whatabout" is the name of the one true god the mindless cults follow.
Re: (Score:2)
But lies of the past had an effect on customers for decades. The revelation that tesla's fsd was faked and actually crashed into a fence could have similar long lasting effects. People do not appreciate being lied to.
Given it's not the only time Musk has faked a demo, I would guess if it has a long-lasting effect it'll be on people's opinions of Musk specifically.
Common sense, where'd you go? (Score:3, Interesting)
These predictions, anyone who has any understanding of computers could have made.
metaverse - That's definitely gonna go away because it was all created in bad faith. VRChat is where any VR "metaverse" experience is happening, and that replaced an earlier experience called "Second Life". The problem with both of these services is that they are not peer-to-peer, and neither is "meta horizons". VRchat is the closest thing there is to "Ready Player One"
cryptocurrencies - that is going to get the hell regulated out of it and it will no longer be viable as governments go after the energy waste.
humanoid robots - Please FFS look up "realdoll X", that is as far as we've been able to get with "humanoid" robots. All it has is a moving face. You know why we don't have stand-up humanoid robots? It's because the necessary weight just to exist is too heavy for things like muscle wire (which gets hot), and other options make too much noise and don't have the articulation necessary. Never mind dexterity. We are not getting a "human sized" human-looking robot until there is an innovation that allows a robot arm or robot leg to be human sized and pull it's own (eg 60kg) weight. These sex-dolls actually have articulation, but in order to not be a "rag doll" they have a heavy steel skeleton so they can sit up in a chair.
language models - We are nowhere close to a general AI. People have been impressed by AI demos, from GPT3/ChatGPT and Dall-E, but these have been cherry picked examples. If you actually use the damn things, they are clearly auto-complete'ing off the input you give it, and aren't particularly good at doing what people have been using them for. No language model will ever be complete enough to use instead of a human, unless the human's entire role was to be a sales person who can't say no, or a customer support chatbot who won't feel emotionally hurt from language hurled at it.
human drivers - Well no duh. The only way human drivers stop being on the road is by upping the drivers licensing requirements to a point that only a robot driver or a really skilled human can pass it. At present, drivers should be re-tested every time they have their license renewed, but that just doesn't happen. Someone who got their license in 1940 can still drive without being retested.
Re: (Score:2)
These sex-dolls actually have articulation, but in order to not be a "rag doll" they have a heavy steel skeleton so they can sit up in a chair.
Steel, really? How sad. Carbon fiber tubing with aluminum joints epoxied to the ends would be plenty durable, anything that broke it would destroy the rubber corpse.
Re: (Score:1)
Plenty durable? Obviously you've never had drunk rage-filled sex with me.
Re:Common sense, where'd you go? (Score:5, Interesting)
We are nowhere close to a general AI. People have been impressed by AI demos, from GPT3/ChatGPT and Dall-E, but these have been cherry picked examples. If you actually use the damn things, they are clearly auto-complete'ing off the input you give it, and aren't particularly good at doing what people have been using them for. No language model will ever be complete enough to use instead of a human, unless the human's entire role was to be a sales person who can't say no, or a customer support chatbot who won't feel emotionally hurt from language hurled at it.
They won't replace humans, but they are going to have a much larger impact than you imagine. The large language models are doing all sorts of things that were impossible a few years ago. Dall-E2, Midjourney and Stable Diffusion are allowing people to make images of a sort easily and efficiently that could not previously be done. The impact there alone is large. And whatever the text models are doing, it is far more than just auto completion. For example, I just asked GPTChat to "Write a short story set in London in the 1870s where magic is starting to come back. Make the main character a young person who is just discovering that they have magical ability." Here's what it responded with, with no retries:
Once upon a time in the bustling city of London, in the year 1870, magic was starting to make a comeback. The streets were filled with the hustle and bustle of people going about their daily lives, but there was a new energy in the air. Some whispered that magic was returning to the world, and that it would change everything.
The main character of our story is a young person named Alice, who had always felt like something was missing in her life. She had always been fascinated by stories of magic and the unknown, but never believed it was possible. But one day, as she was walking through the city, something strange happened. Suddenly, she felt a powerful energy welling up inside of her, and before she knew it, she had conjured a small flame in the palm of her hand.
At first, Alice was terrified. She had never experienced anything like this before, and didn't know what to make of it. But as she began to explore her newfound ability, she realized that she was a witch. She could feel the magic flowing through her veins, and she knew that this was her true calling.
As Alice delved deeper into the world of magic, she met other young people like herself who were also discovering their powers. They formed a tight-knit community, learning from one another and growing stronger together. But as their powers grew, so did the dangers that threatened them.
There were those who feared the return of magic, and they would stop at nothing to eradicate it. They formed secret societies and hunted down anyone who possessed magical abilities. But Alice and her friends refused to be afraid. They stood up to their oppressors and fought to protect the magic that they had come to love.
And so, in the streets of London, a new generation of witches and wizards rose up. They would change the course of history, and ensure that magic would never again be forgotten in the world.
This isn't that great. It is about the level of writing I'd expect from a 12 or 13 year old. But it is obviously not just an auto-complete.
You similarly are underestimating how useful these things might be with only a few tweaks. For example, right now these systems are absolutely awful at doing math, even basic arithmetic. But there is work underway to train them with Coq and Lean, systems which do automated proof checking, so they can learn to output rigorous mathematical proofs that can be automatically checked. At a minimum, physicists and mathematicians will benefit from a system where you can say to it "Here is the definition of a Wibblywobbly. Here is the definition of a Glarglebargle. Prove or disprove that every Wibblywob
Re: Common sense, where'd you go? (Score:2)
Re: (Score:2)
One counter argument would be that all those statistical models work well only when trained on a large body of data, whereas everything we do that matters concerns what has never been done -- we push the boundaries of what is known into the yet unknown to create something new, whether we are a scientist or a baker or a public servant. It follows then that AI/ML could not help you there, they could "only" do better what is mechanical and tedious, which is what computers are for.
Re: (Score:2)
Re: (Score:2)
So maybe it's fair to say AI/ML can discover something new and useful in an area where such a thing can be found and verified through grind. Unlike brute force, the models explore known relations between things to come up with a much smaller space of possible solutions though still large enough that a grind is needed, but what it might take months or years to a person the models can find it in an hour or a day, assuming those relations have been mapped correctly.
Re: (Score:2)
Similarly, and more directly practical, chemists or doctors may be able to train similar systems on databases of chemicals and then ask the systems to output chemicals likely to have specific properties.
A far more significant database to train an AI on is gene to protein folding. This is already in progress and promises bigger steps in genetics than have been seen since the discovery of the double helix. Some of the possibilities are amazing and many are terrifying! But they are going to happen.
Re: (Score:2)
My wife has a master's degree in comp ling and years of professional experience. She could not account for something I got ChatGPT to generate.
"Write a joke" could conceivably be handled by regurgitating training data, but that wouldn't make it funny. I gave it a test on a subject unlikely to have produced humor in the past.
Me: Write a joke about the Treaty of Westphalia.
ChatGPT: Sure!
Here's a joke about the Treaty of Westphalia.
Why was the Treaty of Westphalia a big deal?
Because it ended the Thirty Year's
Re: (Score:2)
These predictions, anyone who has any understanding of computers could have made.
metaverse - That's definitely gonna go away because it was all created in bad faith. VRChat is where any VR "metaverse" experience is happening, and that replaced an earlier experience called "Second Life". The problem with both of these services is that they are not peer-to-peer, and neither is "meta horizons". VRchat is the closest thing there is to "Ready Player One"
I would like to respectfully disagree with you on one point. Just to let you know, Second Life never went away. It still exists and has a hard core group of people who inhabit it (One of which is me). And the fact that it is NOT a peer-to-peer is one of the reasons it still works. To have a virtual world, you have to have a central location (I.E. central server) for the world files to exist upon. That way people can join/leave as they please and the world still exists.
Meta's problem with the "Metaverse" i
Re: (Score:3, Insightful)
Re: Common sense, where'd you go? (Score:1)
Similar thing (Score:4, Interesting)
Just came across this picture [imgur.com] of an old newspaper. What's interesting is a) it's a flip phone which looks remarkably like what Motorola eventually came up with and b) the caption says these types of phones are far in the future. With far in the future meaning roughly a decade [rarehistoricalphotos.com] for implementation and another decade for mass production.
The prediction was right, it just took time to get there.
This is not very useful (Score:2)
some things never change (Score:2)
like typos
Re: (Score:2)
Re: Oh, god, this crap again (Score:2)
Always five years in the future, that metaverse. (Score:2)
Imagine it.
an entire virtual, alternate plane of existence.
Running alongside our own timeline!
We will be able to freely interact with content generated across the world!
People from all over the planet, if they can connect to the Metaverse, all of human knowledge at their fingertips!
But for now we have to settle for less.
We have to settle for a near lightspeed global information network that allows us to do literally everything the metaverses promises.
Bu
Anything is possible (Score:1)
Crypto (Score:2)
Re: (Score:2)
there isa real need for Internet Money that no government or corporation can track or control.
There's no such thing, and there never will be so long as governments exist. A cryptocurrency which cannot be tracked is one which cannot be spent.
Still waiting on flying cars.... (Score:3)