Having signed up the UK already, the data cruncher Palantir is in talks with a number of other European countries who want to use its services in the fight against the coronavirus epidemic. More here.
Today’s Simple AI – AI made easy
And for a bit of fun in these difficult times, have AI artist Al Gahaku turn your photo into a Renaissance masterpiece! Try it out here.
Today’s Simple AI – AI made easy
AI is being rushed out to help with everything from chatbots giving online and telephone diagnoses to logistics crunchers mapping and tracking drugs, masks and machines across entire nations. But is this use of untested algorithms on uncontrolled data sets a help or a hindrance? More here.
Today’s Simple AI – AI made easy TSAI
The EU published this round-up of where AI is being applied in the fight against COVID-19 worldwide. Lots of links to real data scientists working in real time on real problems.
Today’s Simple AI – AI made easy TSAI
The EU is planning the harness the proven ability of AI to quickly spot abnormalities on CT scans to speed up COVID-19 diagnoses. The so-called ‘ground glass’ signature of the virus on the lungs is already being used, but AI could massively cut down on scan time presently required by human experts – from 15 minutes to 10 seconds! More here.
Today’s Simple AI – AI made easy
The Hutter Prize, which rewards the ability to compress data, has been bumped up to a cool half million. Sounds pedestrian in its aim, but they consider this ability a clear marker for AI – consider how neatly we humans fit stuff into our bonces! More here.
Today’s Simple AI – AI made easy
The US gov is using mobile data tracking to check if people are sticking to self-isolation rules and gathering bans. For instance, it was used to shut down a gathering in Central Park. More here.
Today’s Simple AI – AI made easy
Amazon has added some tweaks to help users navigate what steps they need to take if they think they’ve caught Covid-19 – plus it will sing a song while your kids wash their hands! More here.
Today’s Simple AI – AI made easy
The US government is going to release the vast minds of the AI community upon the vast sets of data the COVID-15 virus has thrown up. This could be the beginning of a beautiful friendship! Find out how here.
Today’s Simple AI – AI made easy
Google has launched its real-time translation for Android users (with iOS to follow), meaning you can read simultaneous translation (transliteration from spoken to written) involving 8 languages: English, French, German, Hindi, Portuguese, Russian, Spanish, and Thai. More here.
Today’s Simple AI – AI made easy
Sounds simple, but if we could teach AI to do with proteins what it does with language and music, we could have a key to solving a number of different problem, not least medical ones like COVID-19. More here.
Today’s Simple AI – AI made easy
It’s what we’ve all be hoping for from AI – an all-out blitz on a real global threat. Read here about how scientists, doctors and the AI community are coming together to fight this and future pandemics.
Today’s Simple AI – AI made easy
NeurIPS, a big AI research conference, is asking anyone submitting papers to detail both their financial backing and an estimation of possible positive and negative repercussions of their research. This is a great first step in getting ethical AI into the consciousness of all data scientists. More here.
Today’s Simple AI – AI made easy
Philosophy hasn’t been a job since forever, but suddenly AI firms are finding themselves in moral quagmires and are hiring ethicists to dig them out. But what does the job entail? Find out here.
Today’s Simple AI – AI made easy
The perennial problem of ‘robot stole my job’ could be finding a solution in a USA with record unemployment. It turns out that construction workers don’t mind robots operating backhoes and the like when there’s a huge shortage of skilled workers in their sector. More here.
Today’s Simple AI – AI made easy TSAI
Microsoft has announced that it is making $300,000 available for studies into the ramification of AI for internet and device security. A project can get a maximum of $150,000, so start drawing up those proposals! More here.
Today’s Simple AI – AI made easy
Facial recognition software is being tested by police departments in the States now, which indicates that implementation is not far off. And without too much of a whimper all round. Funny how the body cams to record police behaviour were so controversial… More here.
Today’s Simple AI – AI made easy
A new approach, called the neuro-symbolic model, is trying to get AI to think like a child. Combine some good old-fashioned logic with neural networks, goes the thinking, and you might just get something close to the child’s hard-wired causality and sponge-like receptiveness. More here.
Today’s Simple AI – AI made easy m
Google Assistant can now read a webpage to you in any one of 42 languages, meaning that the AI of translation is coming on in leaps and bounds. Hooray! (hourra, hurra, etc). More here.
Today’s Simple AI – AI made easy
A new computer chip with a light-sensing electronic display and a neural network is being hailed as the first ‘computer eye’. Could it be that just as your last phone had a camera and microphone built in, your next could have an eye? More here.
Today’s Simple AI – AI made easy
One of the big viral sellers in these Coronavirus days is a mask overprinted with your own face – so you can unlock your phone without taking off your mask. Neat idea – but doesn’t this mean your phone is completely vulnerable to anyone with… a picture of you? Come on, facial ID AI – that’s not good enough! More here. Today’s Simple AI – AI made easy
Somehow, our bold and free society suddenly became more tightly surveilled than 1984’s – and for the most part we have passively acquiesced to it when we absently tap that ‘Accept’ or ‘I agree’ button. More here.
Today’s Simple AI – AI made easy
You were so worried about AI taking your job that you didn’t realise it would be becoming your manager! AI is regulating more and more workplaces around the world, from call centres to hotels, and it doesn’t take no for an answer. More here.
Today’s Simple AI – AI made easy
The hacking of Clearview, a facial recognition AI software firm that is used by police departments and police states worldwide, was white hat, done to unmask this photo scraping entity. But in the Wild West of AI and the internet, how long before a black hat gets involved? And isn’t Clearview a bit of a black hat, anyway? More here, here and here.
Today’s Simple AI – AI made easy
And copyrighted them! But don’t despair – it’s for a good cause. In the interests of musical freedom and to stop the burgeoning intellectual property suits, a pair of musicians had an algorithm generate all possible simple melodies, copyrighted the lot, then released them into public domain. More here.
Today’s Simple AI – AI made easy
Joe Redmon, creator of YOLO, has quit the computer vision field because he can’t guarantee his research won’t be used in the cause of evil – or rather is disturbingly sure it will. Chapeau. His Twitter feed with decision here and more here.
Today’s Simple AI – AI made easy
As the progenitors of adversarial attacks pat themselves on the back for having tricked self-driving cars and the like, it’s worth thinking about what a machine deceiving us might look like. More here.
Today’s Simple AI – AI made easy
First use of AI generated deepfakes in a political campaign. This one by the BJP in the Indian election was actually authorized, and has the politician ‘speaking’ a language he cannot, in reality, speak. But it probably won’t stop there, will it? More here.
Today’s Simple AI – AI made easy
A new study indicates that as much as 25% of all internet traffic about climate change is generated by bots – and most of it dresses on the denial side. Sample posts like ‘Get lost Greta!’ may not be high on the sophistication scale, but they pass the Turing test, and are capable of amassing thousands of followers. More here.
Today’s Simple AI – AI made easy
AI did spot the coronavirus outbreak (in late 2019), and did flag it up before people did, but that comes with a caveat – it took humans to sift through all the flags to recognize this as a potential global epidemic. Which makes this the first AI-human disease-busting super-doctor partnership in human-computer history. Read more here.
Today’s Simple AI – AI made easy
An AI machine learning data cruncher has invented a new antibiotic – and it’s wiping out some of the superbugs that had built up an immunity to every other antibiotic known to man. Here is where AI can change the world. Read more here.
Today’s Simple AI – AI made easy
The EU has released its paper on AI and now the Commission’s cogs will soon start whirring. There is no moratorium on face recognition suggested (as an early leak had us believe) but there are some interesting ideas about responsibility, human oversight of black box processes and a law to inform someone whenever they are interacting with an AI entity. Read the report here and an analysis here.
Today’s Simple AI – AI made easy
In more chilling news for driverless cars, the simple application of black tape to a speed limit sign tricked two Teslas into putting their AI feet down – and rocketing ing up from 35 to 85 mph! Where’s an AI traffic cop when you need one? More here.
Today’s Simple AI – AI made easy
OpenAI was founded to ensure that AGI (Artificial General Intelligence) would be a benevolent rather than terrifying master. Their method was to be good guys themselves – non-profit, transparent, etc. But a new article from MTR paints a different picture, and even one of its founders, the Musk himself, says that ALL AI developers should be regulated. Read more here and here.
Today’s Simple AI – AI made easy
A new report from the UK government entitled Artificial Intelligence and Public Standards provides some interesting reading and self-criticism. There’s also 8 tips for public bodies using AI (including full transparency, where it presently falls far short) and 7 tips for AI developers. Read it here.
Today’s Simple AI – AI made easy
In a development which could lead to an ultra-secure internet, inter-galactic travel or the coolest magic show since that Tesla one in the film, a Chinese scientist managed quantum entanglement over the distance of 50km! More here.
Today’s Simple AI – AI made easy
Machines are outperforming human brains on the best stock picks, at least in the long term, according to a new Indiana University study. Which could mean the collapse of the entire financial system – or just go to show quite how bad at their jobs most stock-pickers are. More here.
Today’s Simple AI – AI made easy
The driver who reported the AI self-drive feature in his Tesla had been remotely disabled reported that it has been (remotely) re-enabled – and the switch-off blamed on a ‘miscommunication’. Which is nice, but kind of unease-inducing. Mean to say, we thought when you bought a car, it was yours? More here.
Today’s Simple AI – AI made easy
The WHO has so dubbed the coronovirus, due to the inordinate quantities of both information and misinformation swirling around the interweb about it. It could also be the first time enough data is generated to start some real AI number crunching in the fight against the spread of disease. So good news here or bad news here.
Today’s Simple AI – AI made easy
In a TV moment that can only be described as gut-wrenching, a mother was reunited with her deceased daughter through VR and haptic feedback. Whether this can be therapeutic or not can be left to the experts, because TV is just going plough right ahead! Check it out here.
Today’s Simple AI – AI made easy
With the growing sophistication of phishing attacks (with success rates having grown from 3% to 70%), it is perhaps no surprise that machines are far better at cyber conmen than their human overlords. Time to put AI on the case. More here.
Today’s Simple AI – AI made easy
Facebook may have just advanced the cause of translation AI immensely (not that it was doing so badly already) by releasing a set of over 4.5 billion ‘parallel sentences’ (sentences that mean the same in two languages) from 576 languages. More here.
Today’s Simple AI – AI made easy
In another setback for the AI surveillance software community in the EU, the use of an algorithm to spot possible welfare fraud has been declared illegal in Holland. On the other side of the Atlantic, a New York State high school has started using AI face recognition. Talk about polar opposites! More here and here respectively.
Today’s Simple AI – AI made easy
We keep hearing about racism, sexist and otherwise objectionable AI, which is apparently just ‘born that way’ or something. Here’s an article explaining just how it can happen – and how innocently.
Today’s Simple AI – AI made easy
Turns out the high scores achieved by some AI bots in recognising syntactic differences in sentences might be inflated by a small set of test sentences and the ability of AI to use other clues to make a conclusion. On a larger data set more carefully controlled, AI only fell from 90% to between 59.4 and 79.1% (as compared to humans’ 96%). More here.
Today’s Simple AI – AI made easy
Turns out the high scores achieved by some AI bots in recognizing syntactic differences in sentences might be inflated due to a small set of test sentences and the ability of AI to use other clues to make a conclusion. On a larger data set with more carefully randomized parameters, AI fell from 90% to 59.4-79.1% (as compared to humans’ 96%). More here.
Today’s Simple AI – AI made easy
MIT has released this nifty little bingo game that you can use to help your kid get the concept of AI and deep learning (that’s if they don’t know more than you already!). Suitable for ages 9-14.
Today’s Simple AI – AI made easy
This very upbeat look at the use of drones to warn people to go home or put on their masks during the coronavirus epidemic perhaps hides a little of the nuance behind using drones and AI to control crowds…
Today’s Simple AI – AI made easy
Research is showing that AI which has more of an ‘ego’ is better at making generalizations about the world. Scientists put their AI at dead centre of the game it was playing, and it started getting better results in terms of forming general rules about the game. But does an AI entity with an ego really sound like such a good idea? More here.
Today’s Simple AI – AI made easy
AI’s Meena is outperforming other chatbots by quite a margin, and coming close to matching real people in terms of relevance of response (44-51%, 79% and 86% respectively), plus coming up with some un-terrible humour along the way. (Small question: why do humans only hit 86%?) More here.
Today’s Simple AI – AI made easy
