I wouldn't be too worried about this "theft." That source code is a huge trojan horse anyway, if I ever seen one. How many times has Tesla autopilot caused crashes, deaths, and mayhem so far?
Autopilot is a different code base from the self-driving. Autopilot is a lane keeping and adaptive cruise control that doesn't have anything to do with their self-driving code.
Aye, it may be impossible the way Tesla is trying to do it. Their original plan was for a coast-to-coast demo in 2017, which obviously failed.
What "failed" is that they had to start over from scratch because MobileEye felt that it should own all of the self driving data, and Tesla disagreed. So it took a few years to get back to their 2016 status.
They actually could do a coast to coast demo now and have had that capability for about a year. Their current difficulties are the same that Waymo is having - you have to trust that other drivers will actually obey red lights and stop signs - thus ignoring that the other drivers current velocity will cause a crash if they don't slow down or stop when you make a left turn. Similarly aggressive behavior required for merging, etc. that will cause an accident if the other driver ignores you trying to merge, etc.
Their problem is twofold. First they underestimated the processing power needed to do handle images from the cameras. They use neural nets to process them and on the original hardware they shipped (known as AP2) it just wasn't powerful enough, they couldn't even get it to compare consecutive images (which helps when you don't have stereo vision). They went to AP2.5 and now AP3, but it's not clear if even that is fast enough for what they want to do.
You should watch youtube videos that show the shadowmode debugging output that is tracking people, cars, bikes, road markings (lane boundaries, stop at light boundaries) etc. in real time for all cameras. The hardware works fine for what it needs to do.
The second problem is that it's just really, really hard to use neural nets to do everything they need. Not just recognizing objects like cars, signs and traffic lights. It has to see road markings, it has to see traffic police and understand their gestures, it has to understand complex 3D spaces with no/poor road markings like car parks and private driveways. It has to be able to recognize small objects that the radar/ultrasonics close to the ground won't pick up, like toll barriers and the over-hanging rear ends of trucks.
It isn't as hard as you seem to think, also they aren't using NN's for everything. Also FSD doesn't have to handle every case - you can geofence it - so it never has to handle private driveways. Something that is level 5 for well defined common use cases, but doesn't do country rounds in the middle of nowhere is still a major game changer; or that announces that "in 10 minutes we will be approaching the boundary for FSD, please take over soon".
To give you some idea of how far away they are, even the current driver assist parking isn't good enough for full self driving.
The driver assist is an entirely different code base. It is using essentially none of the data that is being used for FSD development. They are parallel development tracks with almost no resources being devoted to the non FSD stuff.
The early scooter designs had some issues, so Bird changed the design significantly (different battery pack layout is the most obvious), so it is probably just a design changeover, not reflective of the scooter actual durability.
2% of the population are psychopaths; also 2% are sadists.
In a world of nearly 8 billion people - with 4% psychopaths or sadists that is 320 million people who have a natural tendency to do such behavior. So when you are on a platform like twitter or instagram and can be contacted by anyone - chances are you are going to encounter a large number of them. Since the platforms allows some anonymity - they can engage in their behavior with little risk of repercussions.
Not really. There has been a shortage of doctors for decades and it isn't likely to get better. Not everyone is going to rush and train to become a doctor.
Already accredited doctors controlled how many doctors that medical schools were allowed to train and how many new schools were accredited. The supply of students willing to become doctors is enormous and would have greatly exceeded demand - but supply was artificially constrained to ensure that existing doctors could charge higher rates.
So it was actually monopoly control of supply due to artificial constraints rather than students not responding to demand.
I've had an in-person voice conversation with a friend discussing a moderately unknown religious guru. The next day I started seeing ads for their retreat...
Colocation based advertising. You were near the guru location (probably close to an advertised date time and location or an appointment on someones google calaneder ) therefore googles ad network assumed you were interested in the guru.
I'd be curious about long-term testing for these twins if they've developed true permanent HIV immunity and if they're able to pass it along to offspring. Also, if there are any unintended consequences of this modification.
What they are doing is removing a receptor that the HIV virus requires for cellular entry. Since it is being removed at the level of DNA, it should be heritable.
There are known consequences for removal of this receptor - namely greater susceptibility to some other infections such as West Nile Virus; and possibly reduced ability to suppress immune response.
Microsoft had been operating a patent shakedown against open source for a while now. Why is Microsoft doing this? Doesn't add up.
Step 1), transfer all important patents to a patent holding company subsidiary that licenses your patents back to you, and gives you all licensing profits.
Step 2) sign up for patent sharing group, and now that all of your patents aren't "yours", they are not part of the patents required to be shared
Tesla really struggled to ramp up production to meet demand, mostly due to arrogance
They ramped up faster than any car company in history. They are two years from when other car companies would roll their first car off a new line. It is only compared to Musk's hyper ambitious timeline that they 'struggled' to ramp up production.
Since the average age of male physicians and surgeons is 49.1, and the average age of female physicians and surgeons is 42.7 and more recent doctors will be trained better on indentifying female heart attacks - it might be age rather than gender that is giving this result and gender just happens to correlate with age.
It works by targeting mutations of the KAT6A and KAT6B - which are common in a variety of cancers.
When the cells are functioning normally they can be inhibited to put a cell into senescene (sleep) after specific functions are completed, but the mutations result in the these genes being permanently on and never going back to sleep.
Here is a good summary from the end of the nature article,
n summary, using high-throughput screening followed by medicinal chemistry optimization, in-cell assays, biochemical assessment of target engagement and tumour models in mice and fish, we have developed a novel class of inhibitors for a hitherto unexplored category of epigenetic regulators. These inhibitors engage the MYST family of lysine acetyltransferases in primary cells, specifically induce cell cycle exit and senescence, and are effective in preventing the progression of lymphoma in mice.
They didn't note (and don't seem to know) why it doesn't impact health cells but it apparently doesn't based on current testing.
Top secret stuff isn't "50 years ahead" - that is an old estimate and it was related specifically to mathematics of cryptography (NSA was recruiting people in the 1960's and 1970's from pure mathematics fields that the mathematics community didn't figure out applied to cryptography till many years later).
There might be a fairly narrow field that the NSA is ahead, but generally they are probably even with or behind the state of the art - simply because now places like Google, Facebook, etc. have need for the same skillsets.
This is known as word sense disambiguation - there are a number of ways to do so. Training systems for disambiguation is more resource intensive, and in many use cases provides little gain, so most don't bother.
The argument is that the oil companies have knowingly spread false information about climate change - false information that they knew to be true based on their own internal research - resulting in delays in legislation.
So their deception and the damaging results thereof are what the companies are being sued for.
It's not complex cooking at all. Quite frankly I'd like to see them make a robot that could cook steak, potatoes, side veggies meal....because I'm quite certain they couldn't. Even a quality hamburger would confound a robot.
Sous vide is the best way to cook all of those and it is very easily automated.
At first thought, anyway, this "Duplex" thing rather annoys me. If the "person" on whose behalf the bot is calling doesn't feel it is worth their time to speak to me directly, why should I have to waste my time talking to their bot?
This bot makes reservations. As an employee of the company that the reservation is being made at, you wouldn't be "wasting your time" - you would be doing your job.
Why should they? There is no logical reason for them to do so. If the bot works as well in reality as it did in the three demos, thern there is no reason to 'warn' the person on the other end that it is a bot.
Also if the bot can't respond it seamlessly hands off to a call service employee, so there shouldn't be any issues with the bot wasting the time of the reservation takers time.
While Timothy normally does excellent articles, his reasoning and logic were severely flawed this time.
First off the NHTSA report focused on autosteer, not autobraking. Hence his attributing the reduction in accidents to autobraking is bizarre. What the NHTSA was disavowing was that they had not examined the entirety of Autopilot (which includes autosteer, autobraking, lane keeping, etc.). Timothy mistakenly thinks they were stating that they hadn't verified the effectiveness of autosteer installation in accident reduction (they didn't verify the actual usage, but drivers with autosteer installed use it about 50% of their driving time).
Secondly the Tesla's prior to the FSD update already had autobraking, so the 40% reduction in accidents after enabling FSD can't be attributed to the autobraking.
Facebook just trained their image recognition "AI" with over 3 billion instagram images
They then only scored 85% in a test.
That is "top 1" accuracy. A label of a close up of a car might be labeled "fender" but the AI's first guess is car. Or a picture might have a girl holding a cat, and the guess of the AI is cat, but the picture is labeled girl. Or a picture of a racoon, but the AI guesses cat.
These are usually either not actual "wrong" guesses, or their wrogness is fairly minor.
See this article comparing human performance to computer performance for "Top 5".
They are going to rerun the competition vs LeelaZero with the recommendations made by LeelaZero coders (the tuning algorithm that will set it to run max speed on the hardware, and the command line settings that will ensure it runs for the full 50 seconds per move).
From the discussions on the github, it sounds like it should still beat Leela by a significant amount (predictions are 90% of games) though perhaps not as badly as their initial run (100% of games).
Slashdot Mirror
I wouldn't be too worried about this "theft." That source code is a huge trojan horse anyway, if I ever seen one. How many times has Tesla autopilot caused crashes, deaths, and mayhem so far?
Autopilot is a different code base from the self-driving. Autopilot is a lane keeping and adaptive cruise control that doesn't have anything to do with their self-driving code.
Aye, it may be impossible the way Tesla is trying to do it. Their original plan was for a coast-to-coast demo in 2017, which obviously failed.
What "failed" is that they had to start over from scratch because MobileEye felt that it should own all of the self driving data, and Tesla disagreed. So it took a few years to get back to their 2016 status.
They actually could do a coast to coast demo now and have had that capability for about a year. Their current difficulties are the same that Waymo is having - you have to trust that other drivers will actually obey red lights and stop signs - thus ignoring that the other drivers current velocity will cause a crash if they don't slow down or stop when you make a left turn. Similarly aggressive behavior required for merging, etc. that will cause an accident if the other driver ignores you trying to merge, etc.
Their problem is twofold. First they underestimated the processing power needed to do handle images from the cameras. They use neural nets to process them and on the original hardware they shipped (known as AP2) it just wasn't powerful enough, they couldn't even get it to compare consecutive images (which helps when you don't have stereo vision). They went to AP2.5 and now AP3, but it's not clear if even that is fast enough for what they want to do.
You should watch youtube videos that show the shadowmode debugging output that is tracking people, cars, bikes, road markings (lane boundaries, stop at light boundaries) etc. in real time for all cameras. The hardware works fine for what it needs to do.
The second problem is that it's just really, really hard to use neural nets to do everything they need. Not just recognizing objects like cars, signs and traffic lights. It has to see road markings, it has to see traffic police and understand their gestures, it has to understand complex 3D spaces with no/poor road markings like car parks and private driveways. It has to be able to recognize small objects that the radar/ultrasonics close to the ground won't pick up, like toll barriers and the over-hanging rear ends of trucks.
It isn't as hard as you seem to think, also they aren't using NN's for everything. Also FSD doesn't have to handle every case - you can geofence it - so it never has to handle private driveways. Something that is level 5 for well defined common use cases, but doesn't do country rounds in the middle of nowhere is still a major game changer; or that announces that "in 10 minutes we will be approaching the boundary for FSD, please take over soon".
To give you some idea of how far away they are, even the current driver assist parking isn't good enough for full self driving.
The driver assist is an entirely different code base. It is using essentially none of the data that is being used for FSD development. They are parallel development tracks with almost no resources being devoted to the non FSD stuff.
The early scooter designs had some issues, so Bird changed the design significantly (different battery pack layout is the most obvious), so it is probably just a design changeover, not reflective of the scooter actual durability.
2% of the population are psychopaths; also 2% are sadists.
In a world of nearly 8 billion people - with 4% psychopaths or sadists that is 320 million people who have a natural tendency to do such behavior. So when you are on a platform like twitter or instagram and can be contacted by anyone - chances are you are going to encounter a large number of them. Since the platforms allows some anonymity - they can engage in their behavior with little risk of repercussions.
Not really. There has been a shortage of doctors for decades and it isn't likely to get better. Not everyone is going to rush and train to become a doctor.
Already accredited doctors controlled how many doctors that medical schools were allowed to train and how many new schools were accredited. The supply of students willing to become doctors is enormous and would have greatly exceeded demand - but supply was artificially constrained to ensure that existing doctors could charge higher rates.
So it was actually monopoly control of supply due to artificial constraints rather than students not responding to demand.
I believe this 100%. Google Docs has basically become unusable unless you're using Chrome.
It is fairly unusable within Chrome - any document more than a few pages is ridiculously slow to navigate and enter text into.
We're also approaching the year when we were promised self-driving cars. 2018, or ~2017, or 2018. It's going to be a few years of failed predictions.
Waymo started their self driving taxi service today.
I've had an in-person voice conversation with a friend discussing a moderately unknown religious guru. The next day I started seeing ads for their retreat...
Colocation based advertising. You were near the guru location (probably close to an advertised date time and location or an appointment on someones google calaneder ) therefore googles ad network assumed you were interested in the guru.
Requires 150$ for computer time per exam + buying machine time for assignments. So definitely not 'free'.
I'd be curious about long-term testing for these twins if they've developed true permanent HIV immunity and if they're able to pass it along to offspring. Also, if there are any unintended consequences of this modification.
What they are doing is removing a receptor that the HIV virus requires for cellular entry. Since it is being removed at the level of DNA, it should be heritable.
There are known consequences for removal of this receptor - namely greater susceptibility to some other infections such as West Nile Virus; and possibly reduced ability to suppress immune response.
Does this work better than wax applied to apples and other fruit? There is a paper on wax coating avocado's from 1997
http://citeseerx.ist.psu.edu/v...
So I'm not clear what this new company is bringing to the table - does wax (lipids) extracted from peels contain other beneficial compounds.
Microsoft had been operating a patent shakedown against open source for a while now. Why is Microsoft doing this? Doesn't add up.
Step 1), transfer all important patents to a patent holding company subsidiary that licenses your patents back to you, and gives you all licensing profits.
Step 2) sign up for patent sharing group, and now that all of your patents aren't "yours", they are not part of the patents required to be shared
Step 3) PROFIT!
Tesla really struggled to ramp up production to meet demand, mostly due to arrogance
They ramped up faster than any car company in history. They are two years from when other car companies would roll their first car off a new line. It is only compared to Musk's hyper ambitious timeline that they 'struggled' to ramp up production.
This could damage their stock price. I'd sue for defamation - he is making a factual claim as the POTUS on an official communication channel.
Since the average age of male physicians and surgeons is 49.1, and the average age of female physicians and surgeons is 42.7 and more recent doctors will be trained better on indentifying female heart attacks - it might be age rather than gender that is giving this result and gender just happens to correlate with age.
It works by targeting mutations of the KAT6A and KAT6B - which are common in a variety of cancers.
When the cells are functioning normally they can be inhibited to put a cell into senescene (sleep) after specific functions are completed, but the mutations result in the these genes being permanently on and never going back to sleep.
Here is a good summary from the end of the nature article,
n summary, using high-throughput screening followed by medicinal chemistry optimization, in-cell assays, biochemical assessment of target engagement and tumour models in mice and fish, we have developed a novel class of inhibitors for a hitherto unexplored category of epigenetic regulators. These inhibitors engage the MYST family of lysine acetyltransferases in primary cells, specifically induce cell cycle exit and senescence, and are effective in preventing the progression of lymphoma in mice.
They didn't note (and don't seem to know) why it doesn't impact health cells but it apparently doesn't based on current testing.
Top secret stuff isn't "50 years ahead" - that is an old estimate and it was related specifically to mathematics of cryptography (NSA was recruiting people in the 1960's and 1970's from pure mathematics fields that the mathematics community didn't figure out applied to cryptography till many years later).
There might be a fairly narrow field that the NSA is ahead, but generally they are probably even with or behind the state of the art - simply because now places like Google, Facebook, etc. have need for the same skillsets.
This is known as word sense disambiguation - there are a number of ways to do so. Training systems for disambiguation is more resource intensive, and in many use cases provides little gain, so most don't bother.
https://en.wikipedia.org/wiki/...
The argument is that the oil companies have knowingly spread false information about climate change - false information that they knew to be true based on their own internal research - resulting in delays in legislation.
So their deception and the damaging results thereof are what the companies are being sued for.
It's not complex cooking at all. Quite frankly I'd like to see them make a robot that could cook steak, potatoes, side veggies meal....because I'm quite certain they couldn't. Even a quality hamburger would confound a robot.
Sous vide is the best way to cook all of those and it is very easily automated.
At first thought, anyway, this "Duplex" thing rather annoys me. If the "person" on whose behalf the bot is calling doesn't feel it is worth their time to speak to me directly, why should I have to waste my time talking to their bot?
This bot makes reservations. As an employee of the company that the reservation is being made at, you wouldn't be "wasting your time" - you would be doing your job.
Why should they? There is no logical reason for them to do so. If the bot works as well in reality as it did in the three demos, thern there is no reason to 'warn' the person on the other end that it is a bot.
Also if the bot can't respond it seamlessly hands off to a call service employee, so there shouldn't be any issues with the bot wasting the time of the reservation takers time.
While Timothy normally does excellent articles, his reasoning and logic were severely flawed this time.
First off the NHTSA report focused on autosteer, not autobraking. Hence his attributing the reduction in accidents to autobraking is bizarre. What the NHTSA was disavowing was that they had not examined the entirety of Autopilot (which includes autosteer, autobraking, lane keeping, etc.). Timothy mistakenly thinks they were stating that they hadn't verified the effectiveness of autosteer installation in accident reduction (they didn't verify the actual usage, but drivers with autosteer installed use it about 50% of their driving time).
Secondly the Tesla's prior to the FSD update already had autobraking, so the 40% reduction in accidents after enabling FSD can't be attributed to the autobraking.
Facebook just trained their image recognition "AI" with over 3 billion instagram images
They then only scored 85% in a test.
That is "top 1" accuracy. A label of a close up of a car might be labeled "fender" but the AI's first guess is car. Or a picture might have a girl holding a cat, and the guess of the AI is cat, but the picture is labeled girl. Or a picture of a racoon, but the AI guesses cat.
These are usually either not actual "wrong" guesses, or their wrogness is fairly minor.
See this article comparing human performance to computer performance for "Top 5".
http://karpathy.github.io/2014...
They are going to rerun the competition vs LeelaZero with the recommendations made by LeelaZero coders (the tuning algorithm that will set it to run max speed on the hardware, and the command line settings that will ensure it runs for the full 50 seconds per move).
From the discussions on the github, it sounds like it should still beat Leela by a significant amount (predictions are 90% of games) though perhaps not as badly as their initial run (100% of games).