The human-assisted aspect of AI development in general seems to be getting a lot more (well-deserved) attention recently. Google along with everyone else has depended on cheap human labor for data labeling for many years. The fact that Cruise is also relying on human labor to deal with the 2-3% of edge cases the AI can't handle is not surprising. I agree there may be a concern about how it's being represented, and the consequent implications for public perception. It's also true that the cars are operating themselves a significant portion of the time. In my book there is a lot of "self-driving" happening here, just not 100%. But that seems like an unsurprising stage in this process. I have my doubts about whether we will ever get to 100% autonomy, i.e. no human intervention ever, but whether that is a viable goal will be up to the company implementing it to decide... is it more cost-effective to provide their service with mostly-autonomous cars that require occasional human intervention? or is it cheaper just to rely on human drivers in the first place?
Congrats on your launch! Seems like a really slick product.
Have you taken a look at https://cloud.google.com/ml/? I assumed you would have but didn't see anyone mention it in the comments.
Do you see your product as complementary to a service like that? e.g. could you use the Google service as the execution engine for yours? or do you see that as competition?
Not my experience at all. Video quality as far as I can tell is equivalent, and the Skype UX is miles better. Full screen with the little self-view monitoring window at the bottom -- great experience.
My parents-in-law still can't figure out how to get on their G+ account to use Hangouts, but for them Skype is second-nature. Open it, click on the contact, phone rings, presto, we're chatting.
I think the variance on sides is intentional, to balance traffic flow at gas stations.
(Imagine a gas station where most cars use the same entrance and have the tank on the same side. Half of the cars would need to do a U-Turn to use half of the tanks).
I have never seen anybody reach the hose over their car. Not sure why but I'd feel silly trying it. Like 'look at me I don't know which side I should have put the gas in'.
Great point about not hiring jerks, and those are some excellent ideas about how to avoid it. Having worked with, and currently working with, that kind of jerk I've seen how destructive they can be.
I used to live about 8 hours drive from where I'm currently working (rural australia to Sydney). When they found out where I was, they arranged the normal 4 interviews to be arranged back to back rather than spread out over a few days.
Just the fact they were willing to do that meant a lot for me (because no one else had ever bothered how inconvenient it was for me to do interviews).
Still here 3 years later and the attitude shown during the interview process extended right through the company.
I think the comment is just being misunderstood, but Amazon (at least AWS), doesn't do any try before you buy type stuff. It is usually an hour long initial phone screen, than an all day long 7 hour interview with multiple people over the course of a single day in Seattle.
No, it isn't. Unless you're saying you gained weight again while still eating the same diet, you solved the problem, you just failed at implementing the solution.
A diet people consistently can't keep to doesn't constitute a solution to the problem.
It's also worth noting that people who have gained a bunch of weight and lost it have a slower metabolism than people who stayed at the same weight. The Hacker's Diet - being based on calories in/calories out - did not account for this factor. The claim it made was that you could lose weight by keeping a calorie deficit for a while, then once you've lost enough weight you could stop the deficit, go back to basically eating at the level you did before and still stay at the new weight. Which doesn't work.
Dieting seems to work because your body takes a long while to adapt to new conditions. Eventually you get hungrier and less active; progress tends to stop and reverse whether or not you keep "trying".
Temporarily losing weight is "a solved problem". But losing weight without reducing metabolism and thereby making yourself more likely to gain weight in the future, is not.
> But losing weight without reducing metabolism and thereby making yourself more likely to gain weight in the future, is not.
No, it is. Getting people to do it and stick to it is unsolved.
Lifestyle change, to reduce calories and restructure meals to include correct balance of proteins, carbohydrates and fats, and to raise awareness of when food is eaten and how many calories are being eaten, is tricky. But if people do it it works.
You know, if your algorithm worked on the first few test cases you threw at it but after that you kept on finding holes in it, you wouldn't say it had solved the problem, would you?
12 years ago, I lost 55 pounds in six months. All I had to do was carefully monitor everything I ate and burn 1000+ calories exercising six days a week.
Thing is, that was a solution that worked very well for me when I was self-employed 30-year-old single man. It has not worked at all for me as a 42-year-old married man with a child. I don't have the time for 90 minutes of exercise every day. There is always loads of food around I shouldn't be eating. I have vastly more stress in my life than I had then. And my body very definitely reacts differently to food than it used to.
What I'm trying to say is, I found a great hack for losing weight quickly back in the day. I did not find a solution to keep my weight down in the long term.
>>You know, if your algorithm worked on the first few test cases you threw at it but after that you kept on finding holes in it, you wouldn't say it had solved the problem, would you?
The parent said he lost 30 pounds over 6 months. This is not a "test case." He implemented the solution effectively. What he ended up doing however was to change the input variables (literally) by increasing the amount he was eating. Therefore the solution he implemented failed, and he gained the weight right back.
It wasn't sustainable. Maintaining enough of a calorie deficit to lose that much weight made me depressed and made it hard to do my job - I think better with more calorie consumption. So I stopped dieting for much the same reason other people stop smoking - it was having negative effects I didn't like.
(Also, the system I was following (hacker's diet) incorrectly claimed after losing the weight you wouldn't gain it back from simply returning to prior eating habits. Because it didn't account for the now well-known metabolism-lowering effect of dieting.)
The long term solution to keeping your weight down is to eat at your maintenance level. If you do not eat more than your your Total Daily Energy Expenditure (TDEE), you simply will not gain weight. If you eat less than your TDEE, you will lose weight. This might not be easy, but it's pretty simple.
That implies TDEE doesn't change. But we know that TDEE does change when people diet. (One study found that a group of people who lost 10% of their weight through dieting had their resting metabolism decline by an average of 15%.) We don't actually know how TDEE reacts to weight loss in general - how much it declines or for how long it declines in response to a particular level of loss. We also don't know how subjective hunger levels react to weight loss.
...Other than that whatever the formulas are, they seem to make long-term loss nearly impossible for most people.
>(One study found that a group of people who lost 10% of their weight through dieting had their resting metabolism decline by an average of 15%.)
Can you post a link to this study?
>We don't actually know how TDEE reacts to weight loss in general
I don't think this is something "we don't actually know." Lots of weight loss studies have been done in which energy expenditure has been carefully measured, so there's plenty of data on this.
Quote:
"...the majority of the studies point to a reduction in short-term resting metabolic rates that is greater than can be explained by the loss of body mass or fat-free mass over the same time period. Unfortunately, there has been very little work done over the last few years regarding the duration of this phenomenon.
[...]
This is relevant for motivated patients who adhere to severe hypocaloric diets to achieve rather large weight losses. When they get to goal weight their metabolic rate is severely depressed, and they can experience almost immediate weight gain if they resume their prior higher calorie intakes. Recent studies have not continued to measure changes in resting metabolic rate for extended periods to determine whether the reductions are self-limiting. "
The term used these days is "adaptive thermogenesis". You can browbeat it anyhow by sufficiently undercutting caloric intake that the body can't down-regulate enough to cover the deficit. A much safer and easier way to do it is either to use a moderate deficit (studies of athletes show that with identical exercise and identical protein intake, athletes on a 500kcal deficit retain more strength and lean mass than athletes on a 1000kcal deficit) and do exercise, at least some of it with weights.
There's also the "rebound effect", which the Minnesota Starvation Experiment gives us insight into.
People stop dieting and then resume ad libitum eating. They gain fat faster than lean tissue, because gaining fat is easier than gaining lean tissue. Net effect: BF% worsens compared to baseline.
The key is not that "diets don't work, look, they make you fatter"; rather, it's that people see diets as something you do once and then stop. What's actually necessary is ongoing control of food intake.
Ok, let's read this closely. It doesn't seem to me like Larry and David are really denying much of anything.
"Indeed, the U.S. government does not have direct access or a “back door” to the information stored in our data centers."
Sure, that's fine. But the fact that the initial PPT gives a specific date on which Google cooperation began already suggests that Google may be handing over data, not that the gov't is sucking it all down automatically.
Also, they say "data centers" in particular. Data travels over lots of other pipes.
"We had not heard of a program called PRISM until yesterday."
Not surprising the NSA wouldn't tell them the secret code name of the project in their discussions.
"we provide user data to governments only in accordance with the law. Our legal team reviews each and every request, and frequently pushes back when requests are overly broad or don’t follow the correct process."
This is an enormous loophole. "in accordance to the law" could mean anything, and with automated review systems the volume of data passing through could be massive.
"Press reports that suggest that Google is providing open-ended access to our users’ data are false, period."
Not providing "open-ended acces" does not mean not providing any access.
"We were very surprised to learn that such broad orders exist."
But not surprised to learn that other not-so-broad orders exist. How broad is too broad?
"Any suggestion that Google is disclosing information about our users’ Internet activity on such a scale is completely false."
Not "on such a scale", but on a slightly smaller scale, sure.
"there needs to be a more transparent approach"
In other words, we wish we could tell you about everything we're doing with the NSA, but we aren't allowed to.
Really this doesn't read like any sort of denial at all.
