- 1 Hubert Dreyfus
- 2 Uber-Waymo dispute
- 3 Echo Look
- 4 Copying voices with Lyrebird
- 5 Facebook’s Brain-Computer Interface and Elon Musk’s Neuralink
- 6 Opening up of speech and vision APIs to the public
- 7 Viral video of package-sorting robots
- 8 Style transfer
- 9 AI better than doctors
- 10 A terminator for the ISS
- 11 Robots and jobs
- Hubert Dreyfus (1929-2017) has died. He was a man who had still met and talked with Heidegger and Sartre, and yet he belonged equally firmly to the age of computers. He was the greatest and most convincing critic of early symbolic AI, and he used the best of truly humanistic ideas to debunk bad AI, which essentially meant all attempts to make humans into machines, rather than to make machines more human.
Here is a selection of his works (you can support Moral Robots by buying through these affiliate links):
The Uber-Waymo dispute about stolen LIDAR plans goes into the next round. Waymo has, in the public opinion, profited somewhat from the abysmal public image of Uber. But it seems that not everything is so clear-cut about the good and the bad guys in this play. Judge Alsup could not find much evidence that Uber did actually use the stolen plans, and so the whole affair will be around a while longer.
“Echo, look!” Amazon freaked out its customers (and everybody else) this month by presenting a version of its Echo “intelligent” speaker that judges the user’s clothes and that can take selfies by voice control. The worry is, of course, that all these data will leave your bedroom for a round-trip to Amazon’s servers, where they will be processed and digested in order to produce that most disagreeable of digestive products, online advertisements.
Copying voices with Lyrebird
In another piece of freakworthy news, Lyrebird.ai presented their speech-processing system that can ‘style-transfer’ a speaker’s voice and use it to read out an arbitrary piece of text, which the speaker has never seen or uttered before. This will cause all sorts of mayhem with the justice system, as from this moment on no voice recording will ever again be a trustworthy source of evidence. Entirely new forms of fraud will become possible, for example phone conversations with fake versions of one’s friends or family. Good news for children and employees who call in sick, though. Daddy’s voice (“my daughter cannot come to school today”) will finally be available to cover for all misdeeds.
Worrisome news weren’t really in short supply this month. Facebook is researching brain-computer interfaces, and this does not bode well for our freedom of thought, considering Facebook’s way of dealing with all its other issues (advertisements, freedom, hate speech, breastfeeding images, and its fake-Internet cure for world poverty). Accordingly, people were worried, and Facebook plans to put up an ethics committee. The good news: it might be the case that what Facebook wants to do is not even possible.
Brain-Computer interfaces were everywhere this month. Outside of the Facebook debate, Elon Musk also announced Neuralink, his attempt to link brains and computers. In a very insightful piece, Tim Urban (WaitButWhy) brilliantly explains the challenges and dangers of BCI.
Opening up of speech and vision APIs to the public
April saw not only the opening up of flowers, but also of deep learning APIs to the public. Google speech is now available on the Raspberry Pi, and separately people are using the Google vision API to create seeing robots (despite the fact that deep-learning-based vision can easily be fooled). Amazon has similarly opened up the Lex API (the engine behind Alexa and Echo) to experimentation. Good times for hardware-tinkerers worldwide!
Viral video of package-sorting robots
On the lighter side, a video of packaging robots in a Chinese factory went viral, with everybody remarking how cute the little critters that are taking our jobs away are.
If voice transfer (see Lyrebird entry above) is creepy, artistic style transfer is charming (or perhaps also creepy to artists). Finally you can have Van Gogh paint a portrait of your hamster!
AI better than doctors
Human obsolescence progressed further this month, with more evidence that AI is better at medicine than human doctors. Not only can it (March 2017) do better dermatological diagnoses, but now it seems that computers can also predict heart attacks more accurately.
A terminator for the ISS
Since relations with Russia have deteriorated, it doesn’t seem so far-fetched that the Russians would send a killer robot to the ISS to make sure that things are kept tidy and proper up there, and in line with the Superior Chairman’s policies. Unfortunately, it seems that the shooting demo was just an endearing Russian joke, and that the sweet piece of hardware will be delivered to orbit without its handguns. Pity.
Robots and jobs
Finally, there was more discussion on the eternal topic of robots and jobs. Robots to replace 1 in 3 UK jobs over next 20 years, warns IPPR, but also: Amazon creates 1,200 jobs at warehouse equipped with advanced robotics. In a particularly ironic twist, people are now training robots to do their jobs.
With that, it’s probably better that we end here. Thanks for reading, and an exciting May to all!
— Andy@moral-robots.com, http://moral-robots.com