Post ATIyf1AVvtU9DM2F3Q by emilymbender@dair-community.social
(DIR) More posts by emilymbender@dair-community.social
(DIR) Post #ATIyelqospophvhnm4 by emilymbender@dair-community.social
2023-03-05T04:24:34Z
1 likes, 3 repeats
Finally had a moment to read this statement from the FTC and it is 🔥🔥🔥https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-checkA few choice quotes:>>#AIHype #MathyMath #FTC
(DIR) Post #ATIyeo3IhOzUXDYnVQ by emilymbender@dair-community.social
2023-03-05T04:25:26Z
0 likes, 0 repeats
"But one thing is for sure: ["AI" is] a marketing term. Right now it’s a hot one. And at the FTC, one thing we know about hot marketing terms is that some advertisers won’t be able to stop themselves from overusing and abusing them. "https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check>>
(DIR) Post #ATIyeq0tPKFkcKRu1w by emilymbender@dair-community.social
2023-03-05T04:26:04Z
0 likes, 0 repeats
"Your performance claims would be deceptive if they lack scientific support or if they apply only to certain types of users or under certain conditions."https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check>>
(DIR) Post #ATIyernqkn0aASMVyi by emilymbender@dair-community.social
2023-03-05T04:27:04Z
0 likes, 0 repeats
In that connection see:https://twitter.com/emilymbender/status/1632049546527383552And especially:https://twitter.com/timnitGebru/status/1632188168949878784>>
(DIR) Post #ATIyetnDM7gkL452GW by emilymbender@dair-community.social
2023-03-05T04:27:36Z
0 likes, 0 repeats
"Are you promising that your AI product does something better than a non-AI product? [...] You need adequate proof for that kind of comparative claim, too, and if such proof is impossible to get, then don’t make the claim."https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check>>
(DIR) Post #ATIyevigBxFWJZyRTU by emilymbender@dair-community.social
2023-03-05T04:27:58Z
0 likes, 0 repeats
"You need to know about the reasonably foreseeable risks and impact of your AI product before putting it on the market. If something goes wrong – maybe it fails or yields biased results – you can’t just blame a third-party developer of the technology."https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check>>
(DIR) Post #ATIyf1AVvtU9DM2F3Q by emilymbender@dair-community.social
2023-03-05T04:29:31Z
0 likes, 0 repeats
"Does the product actually use AI at all? If you think you can get away with baseless claims that your product is AI-enabled, think again."https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check>>
(DIR) Post #ATIyf3RxS0cwI2DCWO by emilymbender@dair-community.social
2023-03-05T04:30:21Z
0 likes, 0 repeats
That last one is particularly interesting given this from the start of the blog post:"And what exactly is “artificial intelligence” anyway? It’s an ambiguous term with many possible definitions. It often refers to a variety of technological tools and techniques that use computation to perform tasks such as predictions, decisions, or recommendations. But one thing is for sure: it’s a marketing term."https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check>>
(DIR) Post #ATIyf5JAXemk3M7D6G by emilymbender@dair-community.social
2023-03-05T04:30:48Z
0 likes, 0 repeats
I wonder: Would someone who wanted to advertise that their product is "AI powered" or whatevs have to actually define what they mean by "AI"? And how prominent would that definition have to be in their ad copy?>>
(DIR) Post #ATIyf7FLKqug44LBPk by emilymbender@dair-community.social
2023-03-05T04:31:17Z
0 likes, 0 repeats
In sum: I'm really glad to see the FTC putting down this marker, and it gives me hope that there is scope for reining in some of abuses done in the name of building/selling "AI" using existing regulation --- while we work out what further regulation is needed.#AIhype #MathyMath #FTC
(DIR) Post #ATJuAbtlHQmqckbWhU by drsylvainpronovost@sigmoid.social
2023-03-05T13:21:53Z
0 likes, 1 repeats
@emilymbenderGiven that most businesses claiming to do "AI" in-house are in fact simply using multinomial logistic regression, which really is just applied statistical modeling and predates computers (log reg -> 185 years old, softmax -> 155 years old), they have every reason *not* to tell us what they mean by AI...
(DIR) Post #ATJuAdbOwfHxuO1tMO by johnquiggin@mstdn.social
2023-03-05T23:02:35Z
0 likes, 0 repeats
@drsylvainpronovost @emilymbender I've been making this point over and over, but no one seems to get it. It's not an algorithm, it's a model with relatively weak predictive power.
(DIR) Post #ATK3E1rq5FYNcbPUGW by delong@mastodon.social
2023-03-06T00:44:01Z
0 likes, 0 repeats
@johnquiggin @drsylvainpronovost @emilymbender but isn’t that what A I always was? High-dimensional sparse regression, better voice and text interfaces to databases, and the latest way to separate gullible investors in venture capital from their money?
(DIR) Post #ATKDhmGVehKrSgm0XY by patriciajhawkins@mastodon.art
2023-03-06T02:41:27Z
0 likes, 0 repeats
@johnquiggin @drsylvainpronovost @emilymbender Tell 'em it's just predictive text on steroids. And it just finds the next most likely thing to say, with a little randomness thrown in for realism.Most people these days have played predictive text games, and get how bad it can be.
(DIR) Post #ATKRs5XiEgIvVIdA6S by johnquiggin@mstdn.social
2023-03-06T05:20:13Z
0 likes, 0 repeats
@delong @drsylvainpronovost @emilymbender Sure. But if you said “I ran a big computer model using regression analysis to eg predict crimes” people would be sceptical, maybe over sceptical. But call it an AI algorithm and they are totally gullible.
(DIR) Post #ATKSOLy6hVmZiKMTui by patriciajhawkins@mastodon.art
2023-03-06T02:43:12Z
0 likes, 0 repeats
@johnquiggin @drsylvainpronovost @emilymbender If you want to teach somebody something, pick metaphors that are already part of their world.
(DIR) Post #ATKSOMQT04T38Hd87M by johnquiggin@mstdn.social
2023-03-06T05:26:02Z
0 likes, 0 repeats
@patriciajhawkins @drsylvainpronovost @emilymbender Multinomial logit is part of my world
(DIR) Post #ATW1kLhW0ht2OCNrfs by danhon@dan.mastohon.com
2023-03-11T19:24:00Z
0 likes, 0 repeats
@johnquiggin @delong @drsylvainpronovost @emilymbender hell, call it a sophisticated computer algorithm or even sophisticated software and people are still gullible.
(DIR) Post #ATYYnRwKxJibPaK4LQ by patriciajhawkins@mastodon.art
2023-03-11T23:02:09Z
0 likes, 0 repeats
@delong @johnquiggin @drsylvainpronovost @emilymbender Always high-dimensional sparse regression? Well no; in the 1990s, it was fuzzy logic and object-oriented programming (and gullible investors) and before that it was govmn't grants and... oh wait, looky here! https://en.wikipedia.org/wiki/Timeline_of_artificial_intelligenceAnyway, as you say (and IME) it's been a series of big gee-whiz publicity panics that settle down to just being normal niche computer programming techniques.
(DIR) Post #ATYYnSTIyk5d3pkOjQ by delong@mastodon.social
2023-03-11T23:05:54Z
0 likes, 0 repeats
@patriciajhawkins @johnquiggin @drsylvainpronovost @emilymbender so the only constant is gulle investors?
(DIR) Post #ATYYnSxRAiC0ZHqShM by patriciajhawkins@mastodon.art
2023-03-11T23:11:00Z
0 likes, 0 repeats
@delong @johnquiggin @drsylvainpronovost @emilymbender Gullible deep pockets of one sort or another, yes. Though the computing techniques that have resulted are generally solid -- it's just that they're never in retrospect "AI." Or the huge breakthroughs that they're advertised to be.
(DIR) Post #ATYYnTPRUaaty8wpLk by BenRossTransit@mastodon.social
2023-03-12T14:33:07Z
0 likes, 0 repeats
@patriciajhawkins @delong @johnquiggin @drsylvainpronovost @emilymbender I'm old enough to remember when Linear Programming was going to empower computers (IBM mainframes) to run the world.
(DIR) Post #ATYYnU4D31Ci0a1NtQ by johnquiggin@mstdn.social
2023-03-13T00:43:44Z
0 likes, 0 repeats
@BenRossTransit @patriciajhawkins @delong @drsylvainpronovost @emilymbender If you haven't read Red Plenty, you should. Set in Russia when LP was going to make central planning work .