social.stefan-muenz.de

Search

Items tagged with: ai

Yay! 7 ways how artificial intelligence terrifically fails

https://spectrum.ieee.org/ai-failures

#ai #fail
#ai #fail
 
Interview with #TimnitGebru on #AI #regulation starting on the point:

“What can we do right now to make AI more fair?” — “The baseline is labor protection and whistleblower protection and anti-discrimination laws.”

https://www.bloomberg.com/news/articles/2021-09-20/timnit-gebru-former-google-ai-ethics-chief-has-plan-to-rethink-big-tech
 
Tesla AI Day. Yeah, I know, lots of you have already seen the video. So I guess this is for the 3 people who haven't yet.

They think of their AI system as being analogous to the "visual cortex" in biological organisms. The problem they have is fusing the input from multiple cameras. A Tesla car has 8 cameras, which are high dynamic range (HDR) cameras with 1280x960 resolution that operate at 36 frames per second.

The solution to this problem they have opted for is for the neural networks that process the vision to output what they see in the form of 3D vectors, into what they call "vector space" and they can visualize this 3D "vector space" representation on a screen.

The processing first goes through residual networks (resnets), which are convolutional neural networks but the "residual" technique allows them to go much deeper than traditional convolutional networks. They like the fact that they can make the network deeper or shallower as they please to trade off vision processing with latency.

After the resnets, the data goes into something called a BiFPN, which stands for Bi-directional Feature Pyramid Network. They don't say much about what this network outputs, other than that it is "features", not images.

After this the data branches into multiple "heads". Each of the branches does something different: object detection, traffic lights, lane prediction, etc

After this, they do something called "rectification", which takes the vector space output and takes into account each camera's position and orientation and projects its output into the same 3D "vector space". The final fusion process uses a type of neural network called a transformer. These were originally invented for language translation and have an "attention" mechanism that enables the translation system to pay attention do different words in the input as it generates the output. Since then, "vision transformers" have been invented that enable the neural network to focus "attention" on a specific part of a scene. However, Tesla is not using standard vision transformers. They invented their own transformer which operates in "vector space". So it doesn't take images as its input, it takes sets of 3D vectors. What it outputs, at the end of the whole process, is a single unified 3D representation of the scene with curbs, lanes, traffic lights, other cars, pedestrians, and so on, identified.

This system has another trick of its sleeve. Everything up to here is just looking at camera input at a single point in time. But they enabled the system to understand motion over time. This is done with two "cache" systems. One of them is simply time based -- it remembers the last few seconds of whatever the car has seen. The second is space based. So if, for example, the Tesla car sits at a red light, it can remember lane markings it has seen many seconds ago because they are in the "space based" cache and it remembers the space it recently drove past or over.

These "caches" are combined with a recurrent neural network. This combination allows the system to keep track of the structure of the road over time, and the system handles remembering cars when they are temporarily occluded very well.

After all this, the data goes into the planning and control system. For this he shows an example of changing lanes to make a left turn, and says the path planning system does 2,500 path searches in 1.5 milliseconds.

The planning system plans for everything in a scene, including other cars and pedestrians. He shows an example where the car is driving down a narrow street where we can pull aside and yield for another car or they can pull aside and yield for us. If the other car yields, our car knows what to do because it created that plan for the other car.

He shows a visualization of an A* backtracking algorithm, and how it is too computationally expensive and says they are developing a neural network, borrowing the design from AlphaGo, to optimize "Monte Carlo Tree Search", which AlphaGo also does.

You might be surprised that up until this point, the system does not use neural networks, but uses traditional computer science path planning algorithms. In the Q&A section, Elon Musk reveals that these are written in C++. He says neural networks shouldn't be used unless they have to be, and for vision they have to be, but since path planning doesn't have to be it's written in C++.

I would think this system would have trouble working in places with chaotic driving without clear rules, and the presenter acknowledges the system won't work in other places like India, where he himself happens to be from.

Next they talk about data set labelling. Originally they labeled images, but they switched to labeling in 3D vector space. They developed a UI where people can move things in vector space and see the projection in multiple photographs.

He talks about an auto-labeling system, but I didn't really understand how it works. Apparently it can combine data from multiple cars and reconstruct the road surface and walls and other parts of the scene from the video from multiple cars going through the same place. It also does a good job handling occlusions of moving objects such as cars and pedestrians.

They went to the next level by creating a simulator. It makes pretty realistic video. Of course since the simulation is computer-generated the vector space can automatically be correctly labeled and produce massive amounts of training data. The simulation system even simulates the characteristics of the cameras in the cars, such as adding sensor noise and simulating the effect the sun has on the camera. Neural networks are used to enhance the images and make them look even more realistic.

The main purpose of the simulator, though, isn't just to create massive amounts of training data but to create lots of examples of accidents and other edge cases that occur infrequently in real life. Speeding police cars, and so on. Most of the environments are algorithmically created, not created by human artists, so there is a potentially unlimited amount of roads to train from.

Before putting the models in cars, they do extensive testing, with 1 million evaluations/week on every code change. They developed their own debugging tools so you can see the outputs of multiple different revisions of the software side by side.

The rest of the talk is about Dojo, Tesla's upcoming supercomputer.

Basically what they did is create a supercomputer for learning how to drive. They start the process by designing a training node, which is a CPU combined with dedicated hardware for matrix operations (the core operations in any AI system), hardware for parallel floating point and integer math (similar to a DSP chip), SRAM, and communication hardware. The CPU has 4 threads and an instruction set designed specifically for machine learning (so it's not using a general instruction set such as x86 or ARM). 354 of these "training nodes" are manufactured on a single chip, called the D1 chip, with high-speed communication from each node to its adjacent nodes on 4 sides. It has 50 billion transistors on a single 645 millimeter chip manufactured at 7 nm.

With these D1 chips, the plan is to take 500,000 D1 chips and connect them with "Dojo interface processors", which in turn connect to outside computers. The D1 chips are organized into "training tiles". They created their own power supply and cooling systems for these "tiles". The tiles are placed in an "exapod" where 10 cabinets are combined and the walls removed so the tiles can communicate directly with each other without cabinet walls getting in the way.

They made their own compiler to compile PyTorch models and other code for the hardware.

Basically, they created a supercomputer specialized, from the transistors themselves on up, for one specific task, which is training vision neural networks.



#solidstatelife #ai #computervision #autonomousvehicles #tesla
 
Tesla AI Day. Yeah, I know, lots of you have already seen the video. So I guess this is for the 3 people who haven't yet.

They think of their AI system as being analogous to the "visual cortex" in biological organisms. The problem they have is fusing the input from multiple cameras. A Tesla car has 8 cameras, which are high dynamic range (HDR) cameras with 1280x960 resolution that operate at 36 frames per second.

The solution to this problem they have opted for is for the neural networks that process the vision to output what they see in the form of 3D vectors, into what they call "vector space" and they can visualize this 3D "vector space" representation on a screen.

The processing first goes through residual networks (resnets), which are convolutional neural networks but the "residual" technique allows them to go much deeper than traditional convolutional networks. They like the fact that they can make the network deeper or shallower as they please to trade off vision processing with latency.

After the resnets, the data goes into something called a BiFPN, which stands for Bi-directional Feature Pyramid Network. They don't say much about what this network outputs, other than that it is "features", not images.

After this the data branches into multiple "heads". Each of the branches does something different: object detection, traffic lights, lane prediction, etc

After this, they do something called "rectification", which takes the vector space output and takes into account each camera's position and orientation and projects its output into the same 3D "vector space". The final fusion process uses a type of neural network called a transformer. These were originally invented for language translation and have an "attention" mechanism that enables the translation system to pay attention do different words in the input as it generates the output. Since then, "vision transformers" have been invented that enable the neural network to focus "attention" on a specific part of a scene. However, Tesla is not using standard vision transformers. They invented their own transformer which operates in "vector space". So it doesn't take images as its input, it takes sets of 3D vectors. What it outputs, at the end of the whole process, is a single unified 3D representation of the scene with curbs, lanes, traffic lights, other cars, pedestrians, and so on, identified.

This system has another trick of its sleeve. Everything up to here is just looking at camera input at a single point in time. But they enabled the system to understand motion over time. This is done with two "cache" systems. One of them is simply time based -- it remembers the last few seconds of whatever the car has seen. The second is space based. So if, for example, the Tesla car sits at a red light, it can remember lane markings it has seen many seconds ago because they are in the "space based" cache and it remembers the space it recently drove past or over.

These "caches" are combined with a recurrent neural network. This combination allows the system to keep track of the structure of the road over time, and the system handles remembering cars when they are temporarily occluded very well.

After all this, the data goes into the planning and control system. For this he shows an example of changing lanes to make a left turn, and says the path planning system does 2,500 path searches in 1.5 milliseconds.

The planning system plans for everything in a scene, including other cars and pedestrians. He shows an example where the car is driving down a narrow street where we can pull aside and yield for another car or they can pull aside and yield for us. If the other car yields, our car knows what to do because it created that plan for the other car.

He shows a visualization of an A* backtracking algorithm, and how it is too computationally expensive and says they are developing a neural network, borrowing the design from AlphaGo, to optimize "Monte Carlo Tree Search", which AlphaGo also does.

You might be surprised that up until this point, the system does not use neural networks, but uses traditional computer science path planning algorithms. In the Q&A section, Elon Musk reveals that these are written in C++. He says neural networks shouldn't be used unless they have to be, and for vision they have to be, but since path planning doesn't have to be it's written in C++.

I would think this system would have trouble working in places with chaotic driving without clear rules, and the presenter acknowledges the system won't work in other places like India, where he himself happens to be from.

Next they talk about data set labelling. Originally they labeled images, but they switched to labeling in 3D vector space. They developed a UI where people can move things in vector space and see the projection in multiple photographs.

He talks about an auto-labeling system, but I didn't really understand how it works. Apparently it can combine data from multiple cars and reconstruct the road surface and walls and other parts of the scene from the video from multiple cars going through the same place. It also does a good job handling occlusions of moving objects such as cars and pedestrians.

They went to the next level by creating a simulator. It makes pretty realistic video. Of course since the simulation is computer-generated the vector space can automatically be correctly labeled and produce massive amounts of training data. The simulation system even simulates the characteristics of the cameras in the cars, such as adding sensor noise and simulating the effect the sun has on the camera. Neural networks are used to enhance the images and make them look even more realistic.

The main purpose of the simulator, though, isn't just to create massive amounts of training data but to create lots of examples of accidents and other edge cases that occur infrequently in real life. Speeding police cars, and so on. Most of the environments are algorithmically created, not created by human artists, so there is a potentially unlimited amount of roads to train from.

Before putting the models in cars, they do extensive testing, with 1 million evaluations/week on every code change. They developed their own debugging tools so you can see the outputs of multiple different revisions of the software side by side.

The rest of the talk is about Dojo, Tesla's upcoming supercomputer.

Basically what they did is create a supercomputer for learning how to drive. They start the process by designing a training node, which is a CPU combined with dedicated hardware for matrix operations (the core operations in any AI system), hardware for parallel floating point and integer math (similar to a DSP chip), SRAM, and communication hardware. The CPU has 4 threads and an instruction set designed specifically for machine learning (so it's not using a general instruction set such as x86 or ARM). 354 of these "training nodes" are manufactured on a single chip, called the D1 chip, with high-speed communication from each node to its adjacent nodes on 4 sides. It has 50 billion transistors on a single 645 millimeter chip manufactured at 7 nm.

With these D1 chips, the plan is to take 500,000 D1 chips and connect them with "Dojo interface processors", which in turn connect to outside computers. The D1 chips are organized into "training tiles". They created their own power supply and cooling systems for these "tiles". The tiles are placed in an "exapod" where 10 cabinets are combined and the walls removed so the tiles can communicate directly with each other without cabinet walls getting in the way.

They made their own compiler to compile PyTorch models and other code for the hardware.

Basically, they created a supercomputer specialized, from the transistors themselves on up, for one specific task, which is training vision neural networks.



#solidstatelife #ai #computervision #autonomousvehicles #tesla
 

The 'Dead-Internet Theory' Is Wrong but Feels True - The Atlantic

A conspiracy theory spreading online says the whole internet is now fake. It’s ridiculous, but possibly not that ridiculous?
Oh, hey! Yet another one.

#technology #tech #internet #AI #ArtificialIntelligence #bots #ConspiracyTheory
 

The 'Dead-Internet Theory' Is Wrong but Feels True - The Atlantic

A conspiracy theory spreading online says the whole internet is now fake. It’s ridiculous, but possibly not that ridiculous?
Oh, hey! Yet another one.

#technology #tech #internet #AI #ArtificialIntelligence #bots #ConspiracyTheory
 
Datasets. All the famous datasets you've been hearing about from machine learning research for years, collected in one place. CIFAR-10, ImageNet, MNIST, COCO, CIFAR-100, KITTI, SVHN (Street View House Numbers), Cityscapes, CelebA (CelebFaces Attributes Dataset), Fashion-MNIST, Penn Treebank, CUB-200-2011 (Caltech-UCSD Birds-200-2011), UCF101 (UCF101 Human Actions dataset), SQuAD (Stanford Question Answering Dataset), Visual Question Answering (VQA), GLUE (General Language Understanding Evaluation benchmark), ShapeNet, SST (Stanford Sentiment Treebank), LibriSpeech, OpenAI Gym, SNLI (Stanford Natural Language Inference), IMDb Movie Reviews, MuJoCo, miniImageNet, LFW (Labeled Faces in the Wild), ModelNet, Market-1501, Kinetics (Kinetics Human Action Video Dataset), Visual Genome, MovieLens, PUBMED, HMDB51, Places, STL-10 (Self-Taught Learning 10), Places205, MIMIC-III (Medical Information Mart for Intensive Care III), NYUv2 (NYU-Depth V2), BSD (Berkeley Segmentation Dataset), Omniglot, Oxford 102 Flower (102 Category Flower Dataset), LSUN (Large-scale Scene UNderstanding Challenge), Flickr30k, ScanNet, Human3.6M, DBpedia (skiaa), Universal Dependencies, FrameNet, CARLA (Car Learning to Act), and 4,543 more...

Datasets

#solidstatelife #ai #datasets
 
Datasets. All the famous datasets you've been hearing about from machine learning research for years, collected in one place. CIFAR-10, ImageNet, MNIST, COCO, CIFAR-100, KITTI, SVHN (Street View House Numbers), Cityscapes, CelebA (CelebFaces Attributes Dataset), Fashion-MNIST, Penn Treebank, CUB-200-2011 (Caltech-UCSD Birds-200-2011), UCF101 (UCF101 Human Actions dataset), SQuAD (Stanford Question Answering Dataset), Visual Question Answering (VQA), GLUE (General Language Understanding Evaluation benchmark), ShapeNet, SST (Stanford Sentiment Treebank), LibriSpeech, OpenAI Gym, SNLI (Stanford Natural Language Inference), IMDb Movie Reviews, MuJoCo, miniImageNet, LFW (Labeled Faces in the Wild), ModelNet, Market-1501, Kinetics (Kinetics Human Action Video Dataset), Visual Genome, MovieLens, PUBMED, HMDB51, Places, STL-10 (Self-Taught Learning 10), Places205, MIMIC-III (Medical Information Mart for Intensive Care III), NYUv2 (NYU-Depth V2), BSD (Berkeley Segmentation Dataset), Omniglot, Oxford 102 Flower (102 Category Flower Dataset), LSUN (Large-scale Scene UNderstanding Challenge), Flickr30k, ScanNet, Human3.6M, DBpedia (skiaa), Universal Dependencies, FrameNet, CARLA (Car Learning to Act), and 4,543 more...

Datasets

#solidstatelife #ai #datasets
 
Why Teslas keep striking parked firetrucks and police cars. In the opinion of an electrical and computer engineering professor at Carnegie Mellon University, not according to Tesla themselves.

"According to the NHTSA, most of these incidents occurred after dark while the first-responder vehicles were flashing lights and had flares and flashing arrow boards around them. Do you think these lights could confuse the cameras?"

"I’m sure that’s part of the problem. When the lights are spinning and flashing, looking at it from the camera image, these are just pixels, meaning that they have numbers, and the numbers basically go up and down, up and down in some regions, when the light flashes. Unless the training phase has been given those images and that kind of modality, it could throw the pattern matching off."

The CMU professor goes on to say Telsa should use radar or Lidar, but they're not going to do that.

Why Teslas Keep Striking Parked Firetrucks and Police Cars

#solidstatelife #ai #computervision #autonomousvehicles #tesla
 
Why Teslas keep striking parked firetrucks and police cars. In the opinion of an electrical and computer engineering professor at Carnegie Mellon University, not according to Tesla themselves.

"According to the NHTSA, most of these incidents occurred after dark while the first-responder vehicles were flashing lights and had flares and flashing arrow boards around them. Do you think these lights could confuse the cameras?"

"I’m sure that’s part of the problem. When the lights are spinning and flashing, looking at it from the camera image, these are just pixels, meaning that they have numbers, and the numbers basically go up and down, up and down in some regions, when the light flashes. Unless the training phase has been given those images and that kind of modality, it could throw the pattern matching off."

The CMU professor goes on to say Telsa should use radar or Lidar, but they're not going to do that.

Why Teslas Keep Striking Parked Firetrucks and Police Cars

#solidstatelife #ai #computervision #autonomousvehicles #tesla
 
"Xsolla, a company that provides payment processing options for the game industry, has laid off roughly one-third of its workforce after an algorithm employed by the company decided those 150 individuals were 'unengaged and unproductive employees'."

"You received this email because my big data team analyzed your activities in Jira, Confluence, Gmail, chats, documents, dashboards and tagged you as unengaged and unproductive employees. In other words, you were not always present at the workplace when you worked remotely. Many of you might be shocked, but I truly believe that Xsolla is not for you."

Xsolla lays off 150 after an algorithm ruled staff 'unengaged and unproductive'

#solidstatelife #ai
 
"Xsolla, a company that provides payment processing options for the game industry, has laid off roughly one-third of its workforce after an algorithm employed by the company decided those 150 individuals were 'unengaged and unproductive employees'."

"You received this email because my big data team analyzed your activities in Jira, Confluence, Gmail, chats, documents, dashboards and tagged you as unengaged and unproductive employees. In other words, you were not always present at the workplace when you worked remotely. Many of you might be shocked, but I truly believe that Xsolla is not for you."

Xsolla lays off 150 after an algorithm ruled staff 'unengaged and unproductive'

#solidstatelife #ai
 
And now for the latest paper taking the academic AI research community by storm: "How to avoid machine learning pitfalls: a guide for academic researchers". If you haven't heard of this, maybe you're not an academic AI researcher.

Before you start to build models, think about your goals and data. It's important to fully understand the data that will be used to support the goals, and consider any limitations of the data. Make sure the data is from a reliable source, has been collected using a reliable methodology, and is of good quality. Don't look at all your data. This helps reduce the odds you will make untestable assumptions that will later feed into your model. Make sure you have enough data. Talk to domain experts. They can help you to understand which problems are useful to solve. Survey the literature. You're probably not the first person to throw ML at a particular problem domain. Think about how your model will be deployed.

Don't allow test data to leak into the training process. Try out a range of different models. No free lunch. But reporting results from inappropriate models will give reviewers a bad impression of your work. Optimize your model's hyperparameters. But be careful where you optimize hyperparameters and select features.

To robustly evaluate models, use an appropriate test set, and use a validation set. When you train multiple models in succession, using knowledge gained about each model's performance to guide the configuration of the next, you are leaking knowledge from your test set into your models, making it training data without knowing it. A separate validation set should be used to measure performance. You may have to hold out multiple validation sets. You always need some unseen data to evaluate your final model instance. Many models are unstable, so carry out multiple evaluations.

Be careful which metrics you use to evaluate your models. For example, for a classification model accuracy, which is just the percentage of samples that were correctly classified, only works if your classification classes are balanced, otherwise you need to use something else such as Cohen's kappa coefficient or Matthews Correlation Coefficient. Don't assume a bigger number necessarily means a better model. Models trained or evaluated on different partitions of the same data set can give somewhat different numbers. Use statistical tests when comparing models. Correct for multiple comparisons.

Don't always believe results from community benchmarks. Just as how repeatedly making models leaks your test set data into the models, making them implicit training data, so when lots of researchers use the same benchmark, the benchmark data turns into training data. Of course, if access to the test set is unrestricted, you can't assume that people haven't used it as part of the training process, but even if nobody did that, if everyone who uses the data only uses it once as a test set, collectively the test set is being used many times by the community, and the community as a whole is leaking test set data into the models, and some models by chance will over-fit the benchmark.

Consider combinations of models. When reporting your results, be transparent. Share your models and everything people need to repeat your experiment. this will make it easier for other people to build upon your work. Report performance in multiple ways. Don't generalize beyond the data. Be careful when reporting statistical significance. Look inside your models.

How to avoid machine learning pitfalls: a guide for academic researchers

#solidstatelife #ai
 
And now for the latest paper taking the academic AI research community by storm: "How to avoid machine learning pitfalls: a guide for academic researchers". If you haven't heard of this, maybe you're not an academic AI researcher.

Before you start to build models, think about your goals and data. It's important to fully understand the data that will be used to support the goals, and consider any limitations of the data. Make sure the data is from a reliable source, has been collected using a reliable methodology, and is of good quality. Don't look at all your data. This helps reduce the odds you will make untestable assumptions that will later feed into your model. Make sure you have enough data. Talk to domain experts. They can help you to understand which problems are useful to solve. Survey the literature. You're probably not the first person to throw ML at a particular problem domain. Think about how your model will be deployed.

Don't allow test data to leak into the training process. Try out a range of different models. No free lunch. But reporting results from inappropriate models will give reviewers a bad impression of your work. Optimize your model's hyperparameters. But be careful where you optimize hyperparameters and select features.

To robustly evaluate models, use an appropriate test set, and use a validation set. When you train multiple models in succession, using knowledge gained about each model's performance to guide the configuration of the next, you are leaking knowledge from your test set into your models, making it training data without knowing it. A separate validation set should be used to measure performance. You may have to hold out multiple validation sets. You always need some unseen data to evaluate your final model instance. Many models are unstable, so carry out multiple evaluations.

Be careful which metrics you use to evaluate your models. For example, for a classification model accuracy, which is just the percentage of samples that were correctly classified, only works if your classification classes are balanced, otherwise you need to use something else such as Cohen's kappa coefficient or Matthews Correlation Coefficient. Don't assume a bigger number necessarily means a better model. Models trained or evaluated on different partitions of the same data set can give somewhat different numbers. Use statistical tests when comparing models. Correct for multiple comparisons.

Don't always believe results from community benchmarks. Just as how repeatedly making models leaks your test set data into the models, making them implicit training data, so when lots of researchers use the same benchmark, the benchmark data turns into training data. Of course, if access to the test set is unrestricted, you can't assume that people haven't used it as part of the training process, but even if nobody did that, if everyone who uses the data only uses it once as a test set, collectively the test set is being used many times by the community, and the community as a whole is leaking test set data into the models, and some models by chance will over-fit the benchmark.

Consider combinations of models. When reporting your results, be transparent. Share your models and everything people need to repeat your experiment. this will make it easier for other people to build upon your work. Report performance in multiple ways. Don't generalize beyond the data. Be careful when reporting statistical significance. Look inside your models.

How to avoid machine learning pitfalls: a guide for academic researchers

#solidstatelife #ai
 
And now for the latest paper taking the academic AI research community by storm: "How to avoid machine learning pitfalls: a guide for academic researchers". If you haven't heard of this, maybe you're not an academic AI researcher.

Before you start to build models, think about your goals and data. It's important to fully understand the data that will be used to support the goals, and consider any limitations of the data. Make sure the data is from a reliable source, has been collected using a reliable methodology, and is of good quality. Don't look at all your data. This helps reduce the odds you will make untestable assumptions that will later feed into your model. Make sure you have enough data. Talk to domain experts. They can help you to understand which problems are useful to solve. Survey the literature. You're probably not the first person to throw ML at a particular problem domain. Think about how your model will be deployed.

Don't allow test data to leak into the training process. Try out a range of different models. No free lunch. But reporting results from inappropriate models will give reviewers a bad impression of your work. Optimize your model's hyperparameters. But be careful where you optimize hyperparameters and select features.

To robustly evaluate models, use an appropriate test set, and use a validation set. When you train multiple models in succession, using knowledge gained about each model's performance to guide the configuration of the next, you are leaking knowledge from your test set into your models, making it training data without knowing it. A separate validation set should be used to measure performance. You may have to hold out multiple validation sets. You always need some unseen data to evaluate your final model instance. Many models are unstable, so carry out multiple evaluations.

Be careful which metrics you use to evaluate your models. For example, for a classification model accuracy, which is just the percentage of samples that were correctly classified, only works if your classification classes are balanced, otherwise you need to use something else such as Cohen's kappa coefficient or Matthews Correlation Coefficient. Don't assume a bigger number necessarily means a better model. Models trained or evaluated on different partitions of the same data set can give somewhat different numbers. Use statistical tests when comparing models. Correct for multiple comparisons.

Don't always believe results from community benchmarks. Just as how repeatedly making models leaks your test set data into the models, making them implicit training data, so when lots of researchers use the same benchmark, the benchmark data turns into training data. Of course, if access to the test set is unrestricted, you can't assume that people haven't used it as part of the training process, but even if nobody did that, if everyone who uses the data only uses it once as a test set, collectively the test set is being used many times by the community, and the community as a whole is leaking test set data into the models, and some models by chance will over-fit the benchmark.

Consider combinations of models. When reporting your results, be transparent. Share your models and everything people need to repeat your experiment. this will make it easier for other people to build upon your work. Report performance in multiple ways. Don't generalize beyond the data. Be careful when reporting statistical significance. Look inside your models.

How to avoid machine learning pitfalls: a guide for academic researchers

#solidstatelife #ai
 
Nvidia announces TensorRT 8, slashes BERT inference times down to a millisecond - Neowin https://www.neowin.net/news/nvidia-announces-tensorrt-8-slashes-bert-inference-times-down-to-a-millisecond/

#deepLearning #nlp #ai
 

MalwareTech auf Twitter: "Seems it's not just me. Google claims IBM developed TrickBot, Craig Schmugar is the MyDoom author, and Sergey Ulasen wrote Stuxnet. In all three cases the named did researcher on, or discovered said threat, and had nothing to do with authoring or creating it. https://t.co/vvczr6dMq7" / Twitter


#google #ai

https://twitter.com/MalwareTechBlog/status/1415532146593173506
 
Ich finde grad leider keine deutsche version.
#ai #patrick #zaki #bologna #egypt

https://www.amnesty.it/appelli/liberta-per-patrick/

In italien versucht nan grad ihn einzubürgern. Im senat ist es durch, aber das parlament scheint desinteressiert...
Libertà per Patrick
 
Ich finde grad leider keine deutsche version.
#ai #patrick #zaki #bologna #egypt

https://www.amnesty.it/appelli/liberta-per-patrick/

In italien versucht nan grad ihn einzubürgern. Im senat ist es durch, aber das parlament scheint desinteressiert...
Libertà per Patrick
 
"How a largely untested AI algorithm crept into hundreds of hospitals". "The use of algorithms to support clinical decision-making isn't new. But historically, these tools have been put into use only after a rigorous peer review of the raw data and statistical analyses used to develop them. Epic's Deterioration Index, on the other hand, remains proprietary despite its widespread deployment. Although physicians are provided with a list of the variables used to calculate the index and a rough estimate of each variable's impact on the score, we aren't allowed under the hood to evaluate the raw data and calculations.

"Furthermore, the Deterioration Index was not independently validated or peer-reviewed before the tool was rapidly deployed to America's largest healthcare systems."

How a largely untested AI algorithm crept into hundreds of hospitals

#solidstatelife #ai #medicalai #deteriorationindex
 
"How a largely untested AI algorithm crept into hundreds of hospitals". "The use of algorithms to support clinical decision-making isn't new. But historically, these tools have been put into use only after a rigorous peer review of the raw data and statistical analyses used to develop them. Epic's Deterioration Index, on the other hand, remains proprietary despite its widespread deployment. Although physicians are provided with a list of the variables used to calculate the index and a rough estimate of each variable's impact on the score, we aren't allowed under the hood to evaluate the raw data and calculations.

"Furthermore, the Deterioration Index was not independently validated or peer-reviewed before the tool was rapidly deployed to America's largest healthcare systems."

How a largely untested AI algorithm crept into hundreds of hospitals

#solidstatelife #ai #medicalai #deteriorationindex
 
"AI-powered drone deployed in Libya has possibly killed people without any human interference." "The drone, Kargu-2, is made by a Turkish company (STM) and fitted with a payload that explodes once it makes an impact or is in close proximity with its AI-identified target. It is not clear whether the attacks resulted in any deaths."

"The revelations were made in a report published in March 2021 by the United Nations (UN) Panel of Experts on Libya which stated that the drone was a 'lethal autonomous weapon' which had 'hunted down and remotely engaged' soldiers which are believed to have been loyal to Libya's General Khalifa Haftar."

AI-powered drone deployed in Libya has possibly killed people without any human interference

#solidstatelife #ai #robotics #uavs
 
"AI-powered drone deployed in Libya has possibly killed people without any human interference." "The drone, Kargu-2, is made by a Turkish company (STM) and fitted with a payload that explodes once it makes an impact or is in close proximity with its AI-identified target. It is not clear whether the attacks resulted in any deaths."

"The revelations were made in a report published in March 2021 by the United Nations (UN) Panel of Experts on Libya which stated that the drone was a 'lethal autonomous weapon' which had 'hunted down and remotely engaged' soldiers which are believed to have been loyal to Libya's General Khalifa Haftar."

AI-powered drone deployed in Libya has possibly killed people without any human interference

#solidstatelife #ai #robotics #uavs
 

Australian researchers launch AI koala ′facial recognition′ | News | DW | 03.06.2021


I guess they all have the right koalificaitons!
#Australia #koalas #ai
 
Dog poop image dataset for training your robot to pick up dog poop. Apparently it's fake dog poop, too. Was real dog poop that hard to come by?

Dog Poop (MSHIT)

#solidstatelife #ai #datasets
 
Dog poop image dataset for training your robot to pick up dog poop. Apparently it's fake dog poop, too. Was real dog poop that hard to come by?

Dog Poop (MSHIT)

#solidstatelife #ai #datasets
 

Self-Driving or Mind Control? Which Do You Prefer?





#medicalhacks #science #wearablehacks #ai #artificialintelligence #biohack #biopotential #electroencephalography #radiofrequency #hackaday
posted by pod_feeder_v2
Self-Driving or Mind Control? Which Do You Prefer?
 

Self-Driving or Mind Control? Which Do You Prefer?





#medicalhacks #science #wearablehacks #ai #artificialintelligence #biohack #biopotential #electroencephalography #radiofrequency #hackaday
posted by pod_feeder_v2
Self-Driving or Mind Control? Which Do You Prefer?
 
"Most horoscopes ask what month you were born. Co-Star asks what minute. Powered by AI that merges NASA data with the insight of human astrologers."

LOL. Guess what, everybody... "Garbage In, Garbage Out" applies to AI, too.

Co -Star: Hyper-Personalized, Real-Time Horoscopes

#solidstatelife #ai #gigo
 
"Most horoscopes ask what month you were born. Co-Star asks what minute. Powered by AI that merges NASA data with the insight of human astrologers."

LOL. Guess what, everybody... "Garbage In, Garbage Out" applies to AI, too.

Co -Star: Hyper-Personalized, Real-Time Horoscopes

#solidstatelife #ai #gigo
 
"The question of whether machines can think is about as relevant as the question of whether submarines can swim."
  • Edgar Dijkstra
#ai
#ai
 
EU unveils proposals for wide-ranging #AI regulation with a global reach, and facial recognition systems flagged up as “high risk” - https://www.privateinternetaccess.com/blog/eu-unveils-proposals-for-wide-ranging-ai-regulation-with-a-global-reach-and-facial-recognition-systems-flagged-up-as-high-risk/ if it's not watered down by lobbyists, could be as big as #gdpr for #privacy
EU unveils proposals for wide-ranging AI regulation with a global reach, and facial recognition systems flagged up as “high risk”
 
EU unveils proposals for wide-ranging #AI regulation with a global reach, and facial recognition systems flagged up as “high risk” - https://www.privateinternetaccess.com/blog/eu-unveils-proposals-for-wide-ranging-ai-regulation-with-a-global-reach-and-facial-recognition-systems-flagged-up-as-high-risk/ if it's not watered down by lobbyists, could be as big as #gdpr for #privacy
EU unveils proposals for wide-ranging AI regulation with a global reach, and facial recognition systems flagged up as “high risk”
 
Einer der zwei großen #Einzelhändler der #Schweiz, die #Migros spioniert ihre Kunden in deren Filialen nach. Dies mit sogenannten intelligenten #Kamera's und deren #KI / #AI.

Durch #Gesichtserkennung wollen sie dem Ladendiebstahl entlegen wirken doch mittels Gesichtserkennung kann man auch #Kunden tracken und die inkl. ihrer #Emotionen etc.

Es ist keine #Utopie und die #Technik ist da, die politische Debatte über ihren Einsatz aber leider nicht.

#Reclaimyourface

https://www.watson.ch/schweiz/wirtschaft/952083254-geheime-ueberwachungs-kameras-bei-der-migros-firma-geht-gegen-diebe-vor
 
Amnesty International Germany #AI demands immediate release of russian opposition politician Alexej #Nawalny. His life is in imminent danger.
 
Wer mal so richtig Zeit verschwenden möchte… Prokrastination mit #AI
#AI
 
Oddly, in most sci-fi the evil robots hate all humans equally.

#Humour
#AI
 
Oddly, in most sci-fi the evil robots hate all humans equally.

#Humour
#AI
 
"Scientists create online games to show risks of AI emotion recognition." "A team of researchers have created a website -- emojify.info -- where the public can try out emotion recognition systems through their own computer cameras. One game focuses on pulling faces to trick the technology, while another explores how such systems can struggle to read facial expressions in context."

"Alexa Hagerty, project lead and researcher at the University of Cambridge Leverhulme Centre for the Future of Intelligence and the Centre for the Study of Existential Risk, said many people were not aware how common emotion recognition systems were, noting they were employed in situations ranging from job hiring, to customer insight work, airport security, and even education to see if students are engaged or doing their homework."

Scientists create online games to show risks of AI emotion recognition

#solidstatelife #ai #computervision #emotionrecognition
 
Later posts Earlier posts