Kozminski Techblog

Robot legs called Cassie have become one of the new R&D project of Boston Dynamic next to spot, atlas and other fantastic creations. Two-legged robot learned a range of movement from scratch including crouching with a load

Despite the multiple simulations in virtual environments the transfer the models into the real world is a bit tricky, small differences of for example friction can trip over the robot due to the lacking data from simulation. when a robot tries to apply what it has learned. A heavy two-legged robot can lose balance and fall if its movements are even a tiny bit off.

In this case Boston dynamics applied double simulation where the Cassie learned from the virtual and real environment. The data collected from the real world was transferred to environment called SimMechanics that mirrors real-world physics with high degree of accuracy

The real Cassie was able to walk using the model learned in simulation without any extra fine-tuning. It could walk across rough and slippery terrain, carry unexpected loads, and recover from being pushed. During testing, Cassie also damaged two motors in its right leg but was able to adjust its movements to compensate.

Refference:

https://www.technologyreview.com/2021/04/08/1022176/boston-dynamics-cassie-robot-walk-reinforcement-learning-ai/

April 2021 India is under huge new wave of novel corona virus. The hospitals are lacking in equipment and staff. Around social media people are tracking down oxygen, drugs and lacking medical equipment. Government is almost helpless with occurring situation.
Sohini Chattopadhyay 30, with a group of her friends were struggling to finds donors of plasma on all sort of social medias with no results.
Chattopadhyay finally decided to use tinder to find the donors of the blood.

She and another friend of hers created profiles each on the dating service and were swiping right of anyone who looked healthy and was close to their age. they managed to find the donor with the right blood type who was willing to donate the plasma.

“I was pretty touched,” says Chattopadhyay. “What started as a desperate joke turned out to actually give us some leads, and in this case also found someone willing to donate.” 

Although a Tinder match for convalescent plasma may be a one-off success story, it represents the myriad ways India’s residents are going online to help their loved ones as the country faces around 350 000 new covid cases a day, a dramatic spike that began in early April. And without enough government action or information, ordinary citizens are turning to social media to crowdsource everything from financial help to medical equipment. They’re inundating Twitter and Instagram with requests for hospital beds, oxygen supplies, antiviral drugs, and convalescent plasma donors; they’re creating Google docs, websites, and web apps to aggregate what’s being shared and to play matchmaker between buyers and sellers. 



Reference:
https://www.technologyreview.com/2021/04/28/1023983/india-covid-crowdsourcing/

Reading Time: < 1 minuteRecently Nvidia (one of the biggest chip producers) created a project using GANs neural network.

Basically, the programme generates new faces of people, who never existed. It uses data which was fed initially into it and uses it to create a new face. The face features elements of chosen age, skin color or even details like freckles. The video below shows how the faces are generated using sample photos.

Personally I’m amazed with the technological progress which happened within the last 3 years. This technology will definitly be used in video games in the upcoming years. What’s more, the use of GANs neural networks is still at its very beginning but it will have more usage with time. I think it will be used for example to replace interior decorators in the spectrum of the next 5 years. On the other hand it will also boost the quality of deep fakes and will definitly be used to manipulate data. What do you think? Let me know what are your thoughts?Personally i’m amazed with the technological progres which happend within last 3 years. This technology will definitly be used in video games in few years, What’s more the use of GANs neural networks is still at its very begining but it will have more usage with time, for example i think it will be used to replace interior decarators in the time of 5 years. On the other hand it will also boost the quality of deep fakes with will definitly be used to manipulete data.

What do you think let me know.

refference:

https://www.technologyreview.com/s/612612/these-incredibly-real-fake-faces-show-how-algorithms-can-now-mess-with-us/?utm_campaign=the_algorithm.unpaid.engagement&utm_source=hs_email&utm_medium=email&_hsenc=p2ANqtz-8kxzmyRmbG-8kGLZmbD-jZcUEpJidf6MnVsJOK62S0aVm88ioDTuQC2FPFAnkZQDBiqBw7

Reading Time: < 1 minuteWhat are deep fakes and what will be their future

Deepfake is a combination of word deep learning and fake. Basically deepfake are videos which use already existing image to combine with a video just like the one below with Nicolas Cage. Mainly deepfakes were used to create fake news and hoaxes.

Right now even in initial state of deepfakes it is hard do distinguish real video from the fake one. In the future it will be even harder with development of this tech and huge amount of data available in the web. For example fake videos of celebrities or politicians already exists which are almost perfect. Despite the video itself but also audio and specific facial movements of person can be “trained” if there is enough data. the video below shows previous US president Barack Obama

It only shows that we already need to double check crucial information because there are people who finds joy in creating false visions. What’s more deepfake’s algorithms can be used other way around to check videos if there was any interference of AI.
This technology on the other hand can help upgrading video chats it uses much less data that actual streaming. This for me is only positive aspect of usage deepfakes right now, with good timing this tech can destroy numerous peoples’ lives, just like it happened with Gal Gadot whose face was used in porn video.

REFERECE:

The future of the deepfake — and what it means for fact-checkers

https://en.wikipedia.org/wiki/Deepfake

Reading Time: 2 minutes

About 2 years ago it was the first time when I heard about AI. Now it seams that it could help radiologists to check X-rays . Back then it wasn’t even in Alpha tests and now it  becomes reality.

Right now in Europe when mammograms are performed, normally 2 radiologists are looking for the signs of cancer. In a near future it will most certainly change.  AI will be detecting irregularities in x-rays scans thanks to deep learning and big data base of images of breasts with anomalies.

A few start-ups already started working on those programs e.g. Berlin-based MX Healthcare which collected over 1 million imagines (just like the one below on the left). Google also started working on AI  with the same purpose using program called Lymph Node Assistant, or LYNA.

Above: Left: a slide containing lymph nodes. Right: LYNA identifying the tumor region.

Image Credit: Google

LYNA detects metastatic tumors made of cancerous cells which break away from their original tissue , circulate through the body and form new tumors in other parts of the body. Those cells are extremely difficult to detects and due to  that there are around 0,5 millions deaths annually. In one of the tests Google operating system received 99% accuracy ( comparing to human with only 62% ) which only tells that in future AI will be used 100% with or without help of radiologists.

The difference between LYNA and other AI is a data base. Google AI uses open source data which let it learn and practice more efficient than other programs.

To sum it up, AI in the future will be checking probably all X-ray imagines due to it superiority over humans, but it doesn’t mean that radiologists will lose their jobs. This is because AI will be able to check the past of the cancerous changes but not the future of them.

References:

1. https://venturebeat.com/2018/10/12/google-ai-claims-99-accuracy-in-metastatic-breast-cancer-detection/

2. https://www.politico.eu/article/ai-doctors-cancer-treatment-deep-learning/?utm_source=POLITICO.EU&utm_campaign=1107d30718-EMAIL_CAMPAIGN_2018_11_23_05_18&utm_medium=email&utm_term=0_10959edeb5-1107d30718-190318285

3. https://www.japantimes.co.jp/news/2017/10/25/national/science-health/cancer-breast-density-doctors-withholding/#.XAQs02hKg2w