Prety neat driving
But the car seems to be a long way away from the kerb. Maybe that's standard for the US?
Watch Waymo's totally driverless self-driving car cruise around, how the US military wants to use AI ethically, etc
Hello, here’s a short but sweet round up of news from the world of machine learning beyond what we have already covered on El Reg. Microsoft funded an AI startup that spies on Palestinians: Microsoft invested in AnyVision, a company that supports a secret Israeli military project that surveils Palestinians travelling within …
-
-
Monday 4th November 2019 11:09 GMT stiine
Re: Prety neat driving
The U.S. tends to put bike lanes on the right, and they're about 3' wide.
On the other hand, driving closer to the center line (and depending on locale and conditions, over the center line) is safer because the more distance you put between your vehicle and the pedestrians*, the fewer evasive maneuvers you have to perform.
* - or wildlife.
-
Monday 4th November 2019 14:07 GMT Charlie Clark
Re: Prety neat driving
In many of the places in the US I've been to there are no pavements and there seems to be little distinction between them and wildlife.
Maybe it's just the perspective of the camera but it seemed quite a bit further into the road than in Europe. Then again, here, you'd generally expect to move to overtake cyclists.
-
Monday 4th November 2019 18:41 GMT Anonymous Coward
Re: Prety neat driving
No pavements is a thing in other countries too, like Japan, for instance. And actually the latest theories on traffic management in urban areas state that pavements are bad anyway because they make drivers complacent. If only there were no clear demarcation between vehicles and pedestrians then drivers would need to take much more care and drive more slowly. Yeah, right.
-
-
-
Monday 4th November 2019 12:01 GMT MiguelC
Re: Prety neat driving
TechCrunch got to ride in one of those... pretty boring if you go by their account
-
Monday 4th November 2019 16:00 GMT Alister
Re: Prety neat driving
It was interesting to see that when the Waymo car came up to an intersection and signalled left, and then the following car pulled alongside on its left, there was a detectable pause whilst the Waymo car tried to decide what to do - you could see the steering wheel twitch back and forth for a second or so.
-
Monday 4th November 2019 09:58 GMT Il'Geller
The uniquenesses.
Google "quantum" computer, Deepmind's AlphaStar AI bot, Waymo are the direct result of AI-indexing and creating synonymous clusters.
1. "Extended", "prolonged" dictionary definitions on patterns' words are created. That is, most of words (in dictionary) have synonyms, and these definitions are chained one to another (as paragraphs), and thus chains (approximately 10 bytes per 2-5 words) are used, which are filtered through the surrounding contexts and subtexts. The idea is to create the true uniquenesses: after the formation of synonymous clusters these uniquenesses convey the true essence of what is said, and can be instantly found.
2. The uniquenesses are used to form synonym clusters, which replace all programming languages commands.
3. Thus one does not need programmers and can immediately apply direct speech, as AI.
In NIST TREC I had to find one unique phrase in over 6 million texts, which I did. IBM Watson in Jeopardy! had to find one unique phrase in millions of texts, which was done.
How? By the uniquenesses.
-
Monday 4th November 2019 13:34 GMT Il'Geller
Re: The uniquenesses.
A unique action can be assigned to a group of patterns. For example, a group is received from Waymo's sensors (an old woman crosses the street), and this description is opposed to the action of "braking". So, the whole problem boils down to
1. Descriptions on everything what happens (labels, annotations) in the form of texts are needed.
2. Removal from texts of lexical noise is the must; where the noise is typically superfluous patterns, which do not explain the central themes contained within the texts and, accordingly, removal of it results in an improvement in the quality of the structured texts.
3. Creating a synonymous cluster is needed as well.
5. Assigning of synonymous clusters to certain mechanical actions.
Savings in this case is due to the cost of programmers.
-
-
Monday 4th November 2019 18:12 GMT Il'Geller
4. Blockchain.
4. Establishing a time hierarchy between synonymous clusters, the so-called blockchain technology, to identify cause-and-effect relationships.
For example, for somebody, who already ate a whole cake, makes no sense to advertise a new cake, but rather a simulator for the rapid weight loss. This is a cause-and-effect relationship, as a blockchain between synonymous clusters, which are aggregated by meanings together (into paragraphs).
-
-
Tuesday 5th November 2019 08:06 GMT Il'Geller
Yes, it's possible.
It's AI, this is a personalized technology, it's you with all your guts in cyberspace. That is, the computer over time becomes more and more you, through remembering how and what synonymous patterns are associated with what timestamps in your life. And then the choice of advertising (which you see) is dictated by these timestamps: if there is a temporal correlation of your trips to the gym with your gluttony, then advertising for sports equipment gets priority and should come out on the screen of your computer.
THIS is happening now as well, but now you are owned by the spying Google and FB. AI technology allows you to own yourself, as profiling is now possible offline.
-
Tuesday 5th November 2019 13:44 GMT Il'Geller
Timestamps and chronology (for synonymous clusters?)
I am not sure that Google (Waymo) uses timestamps and chronology for synonymous clusters... While they can make driving much easier, because they allow a car's AI to anticipate possible difficulties, being ruled by what DeepMind calls "scenarios". That is, seeing an old woman on the road the AI can call from its memory all the scenarios, at once (uniqueness!) select one according to the information gotten from its sensors, and make its decision. For example to rotate right and avoid the lady, rather than to slow down.
However, DeepMind uses scenarios, doesn't it?.. And Google said something completely indigestible and inarticulate about its alleged "quantum" computer... So, Google marks patterns (synonymous clusters?) by timestamps. Impossible to work when you can only guess!
-
-
-
-
-
-
Monday 4th November 2019 14:23 GMT Charlie Clark
Re: Did any one else notice
Yes, we all noticed the obvious. NB. the wasn't a demonstration video, just some footage from an interested driver. We don't know anything about the journey itself and what it may or may not have been testing.
Google has chosen Phoenix for several reasons: cooperative state and council and predictable roads and weather. As for hostile driving environments: forget London, and pick anywhere in South Asia or Africa.
-
-
Monday 4th November 2019 23:19 GMT teknopaul
Re: Did any one else notice
A car jackers dream.
If they cover the cameras it will probably stop on its own. They pop it in a box and off to the chop shop, while advanced AI tries to work out why visibility < 0.02m && elevation is 0.2m > max_shock_extension.
The kids are gonna have so much fun with them things. I'm sure if you jump out in front of it at just the right place you can get one to drive into the duck pond.
-
Tuesday 5th November 2019 09:31 GMT Charlie Clark
Re: Did any one else notice
Not really, the car can be built to be largely useless if hijacked and the components are worth a lot less than car manufacturers would want us to believe. The cars belong to Google so they can be remotely bricked and, of course, Google gets to decide whether somewhere is safe for their cars to go.
-
-
Tuesday 5th November 2019 15:22 GMT Charlie Clark
Re: Did any one else notice
This particular model maybe, yes, but I wouldn't expect that to be the case always.
There will no doubt be attacks on the cars but it's difficult to see why there should be of these already and that Google will be powerless in such cases. Around here car theft is already mainly limited to stealing the sat-nav systems because these have the highest resale values. Stealing a car just for its scrap value, which is what you're suggesting, is unlikely to make the perpertrators rich, and, as the owner Google can install and activate all kinds of disincentives that car manufacturers normally can't.
Vandalism (the yoof will often trash anything whatever its worth) may be another issue, but again, Google can quickly cut its losses and, presumably, get good at working out where such things are likely to occur: in the true tradition of the internet, I'd expect the engineers to see such hazards no differently to difficult driving conditions.
But I guess, as with so many things, we'll just have to wait and see.
-
-
-
-
-
-
Monday 4th November 2019 14:40 GMT Charlie Clark
Re: US Department of Defense has drafted its own AI principles
Betrayal, particularly of the Kurds, is one of the few constants of the Middle East. Sykes-Picot being a notable example of Anglo-French cooperation in screwing the Kurds but there have been many others. However, in this particular case the DoD was dead against the decison which was the Orange One's alone, who also only knows why he gave the order: to please Putin, to deflect from impeachment hearings, to shore up his base, because he was bored.
It's not that US hasn't abandoned allies in the past, but it rarely does so with so little upside: Erdogan gets to like a hero, Putin a statesman, Assad gets some oil revenue and Iran gets a little close to Israel.
US intervention in Syria had thus far been relatively cheap (both in men and money) and reasonably successful. Compare and contrast with the morass that is Afghanistan.
-
Tuesday 5th November 2019 00:45 GMT Clunking Fist
Re: US Department of Defense has drafted its own AI principles
"Orange One's alone, who also only knows why he gave the order"
Because he promised to take the US out of foreign wars? It laugh out loudable how much "liberals" and "progressives" warmonger and want the US to defend one bunch of rogues over another. The "liberal" media were aghast when he declined to attack Iran. And were quite quiet when Obama destroyed Libya...
Remember, real people die in these wars.
Not since World War 2, has any war involving the US ever come to a conclusion. Trump (and true liberals) decided that it's best not to get involved in the first place...
He's a bit of a twat, but at least he's not a warmonger. He may even prevent the US getting more involved in Venezuela..
-
-
Tuesday 5th November 2019 09:45 GMT Charlie Clark
Re: US Department of Defense has drafted its own AI principles
IIRC the stated objective was to destroy IS/DAESH.
US foreign policy, including all those expensive wars, is largely dictated by economic concerns: if others hadn't already done so the US would have invented gunboat diplomacy as yet another expression of the Monroe Doctrine, when Coca-Cola diplomacy fails.
"Rebuilding Americas Defences" made the oil-based case for invading Iraq long before the event.
Syria is a bit more compicated because it has less oil but it does have a border with Israel, not just an ally but a useful R&D site for some of DARPA's more dubious projects.
Trump doesn't care about American lives, but he does care about winning elections: any blowback from Syria is unlikely to hit the US before next year's election.
-
Thursday 7th November 2019 01:49 GMT Clunking Fist
Re: US Department of Defense has drafted its own AI principles
""Rebuilding Americas Defences" made the oil-based case for invading Iraq long before the event."
Aye, we all know that WMD was an excuse, was fake news. Just like "chemical attack" in Syria is...
"Trump doesn't care about American lives, but he does care about winning elections: any blowback from Syria is unlikely to hit the US before next year's election."
Lol, that's known a Trump derangement syndrome. What makes you think that he doesn't care about American lives?
As for blowback, the media are the only ones who can give blowback. There's a chance they are planning to do it in the build-up to the election. However, Hillary is still yet to receive blowback for the destruction of Libya, for the death of the US embassy staff in Benghazi.
Fracking in the USA should help save lives in the middle east. The swamp doesn't seem interested...
-
-
-
-
-
Monday 4th November 2019 20:45 GMT Dinanziame
Somewhere in Phoenix, or rather Chandler
S Welch – W Pelican – South Dakota – W Flamingo
-
This post has been deleted by its author
-
-
-
Monday 11th November 2019 10:29 GMT froberts2
It doesn't matter. Any even slightly competent driver would have NOT turned left at the point that the other driver had pulled along side of them. The obvious assumption is that the 2nd driver had not realized that the lead car was turning left (even though it had indicated that fact), but was instead turning right.
The alternative assumption is that the 2nd driver was acting aggressively and attempting to overtake on the turn.
For either of those conclusions, proceeding to turn left at that point, especially after only a short wait, is hazardous and almost certainly the wrong action to take.
-
-
