Ai-Da's responses were recorded from a human actor, not actually generated by the robot at all.

Dead giveaway: at 1:43, "for my poetry... using neutral networks". I find it very much easier to believe in an actor saying "neutral" for "neural" than an AI doing it.

Although I suppose it is possible that it was a typo in the script that the robot is reading. But that would mean that the intonation is really very impressive.

> If Ai-Da's art skills are anything like her speaking performance at the hearing, they is probably quite atrocious

Did an AI write that sentence?

Still, members of the House of Lords appeared to take Ai-Da at face value.

Besides the bad pun(s), the non-techies are impressed with tech showing binkenlights, movement and sound. Magic still is the most convincing presentation.

I'm waiting for the alien invasion. That must be better than seeing the political establishment's bedazzlement with magic.

I normally can't stand podcasts but this fake podcast is amazing .

and for the faddists, scammers, and the hype machine to move on to the Next Big Thing, and for AI to be relegated back to the few things it can do safely and well. (I'm assuming there are a ffew).

Remember when Prolog was the Next Big Thiing and was going to save/revolutionize the world?

Remember when expert systems were the Next Big thing and were going to save/revolutionize the world?

etc.

I saw this, on bbc rather than the guardian, earlier this week, and I found it fascinating. I mean, on one side, laughable, to watch all those grannies, etc., getting excited about a speaking doll, on the other hand, it was not some wacky 'future's here' demo geek conference or show, it's whether I like it or not - a place where decisions are made about this country, and about my life too. Though, really, I still think it was a waste of time and a gimmick for the grannies' sake. I mean, do we not have _real_ problems to solve now? Though, I guess, getting a nice extra stream of financial support from the treasury wouldn't hurt any business, eh.

Right. As opposed to being created by an experienced robotics or software engineer, it was created by someone who runs an art gallery. Who will know plenty of artists who can knock up a mildy-plausible looking robot. Which can then be fitted with a basic program that uses speech recognition and synthesis to answer a limited number of known questions with preprogrammed answers.

Unsurprisingly, despite the demo crash, this was enough for all the Peers of the Realm to pronounce themselves dazzled by this amazing new machine, and feel reassured that Britain will soon be at the forefront of the imminent AI industrial revolution.

If this is all it takes, there must be huge opportunities available to knock up something similar in your garden shed, present it to the politicos as an exciting AI prototype, and bag big bucks for "further development and research".

I thought the weird robot that crashed in Parliament was called Liz Truss.

Interesting presentation.

One of the things humans are very bad at is in being set in our ways, holding out-dated views, etc. Off the top of my head, Daylight Saving is a good example. Why is it still a thing? Ok, that is a rhetorical question for the purposes of my post.

One area where AI might excel is in acting as a pre-processor for - for example - government policy. It is not there to be the final aribiter of choice, but to signpost unintended consequences and, in control theory terms, to prevent any self-reinforcing problems which produce a positive feedback loop - thinking here of Putin's thought processes, for example (far-fetched, I know, but the human mind often takes one down a rabbit hole). The output from the AI system would then be used to make an enlightened final decision.

This concept has been around for years. When the (London Undergound) Watford computer pilot was being commissioned the designers ran the track layout through all possible valid signalling moves and discovered one that nobody had thought of. So where do you draw the line between simple data analysis and AI? My guess would be in the production of iterations of data, the results of which would feed back into the datastore. So how do your protect your datastore from going down the same rabbit hole that Putin has conjured? Somehow the brain discards options from decision-making which are perfectly ok, and chooses an option that leads to scorched earth. There might be no way to get an AI system to learn that that is not an option, so it will have to be manually selected by a human, and scored with a heavy negative weight. Once you intervene in this way, is it still AI?

My negative feelings towards AI are unchanged (AI cannot be left alone to make decisions, it must be monitored, and the rationale behind any output must be capable of being unravelled).

Recently I had a problem with a purchase I made on line. I used the on-line chat bot to make a complaint. I had to have a few go's at formulating a question that the bot could understand, so straight away I could tell it was a bot response. Then there were a few symptoms displayed, to which I chose accordingly. The problem with these systems is that they are there to completely automate customer service. If you fall out the bottom without resolution of the issue then the ultimate question is "Would you like a refund?" There was no option to speak to a human being. There are companies out there who can't be arsed to deal with problems, the loss of one sale is but one drop in a massive ocean, so they don't care. This is a fad that everyone seems to want to use because it's sexy.

As regards the presentation, Ai-Da's minder insisted that "she" be given a paint palette and brush to create output. The traditional metaphor for an artist. But that metaphor is already an out-dated concept with (for example) David Hockney producing some very Hockneyesque pictures using a digital camera to produce a photo montage, plus his work with tablet computers.

Largely because if Ai-Da had had any contact with Tay or similar beforehand, the output might deviate from the script and be considered somewhat uncouth.

The other reason is surely the size of Ai-Da's data repository. How big is it? To do paintings, etc. surely we must be looking at petabytes?

"a version of Designer will be integrated into the Edge browser at some point."

So how will MS get people to use it?