Ancestors

Toot

Written by Martin Gleadow on 2024-11-18 at 19:05

Anyone using AI without a sound, validated and approved use case should be held accountable for their reckless use of finite resources for pointless purpose

=> More informations about this toot | More toots from mgleadow@mastodon.green

Descendants

Written by Will Deakin on 2024-11-18 at 21:21

@mgleadow My issue with AI discourse is what is meant by AI?

For example, the Dijkstra and A* algorithms were "AI" when they were developed. They or a variant are arguably used in all route finding applications today.

Or, if I build a machine-learning model using population density, the centre-line track-model and historic rail passenger data to predict travel likelihood. It uses scaling, a Histogram Gradient Boosting regressor, cross-validation and so on. So feels AI. But is it wrong?

1/n

=> More informations about this toot | More toots from wnd@fosstodon.org

Written by Martin Gleadow on 2024-11-18 at 21:24

@wnd yes, I should have been more precise in use of language. I was specifically referring to the generation of unreliable filler content through the regurgitation of hastily hoarded input without consideration to nuance.

The use of scientific approaches to definable problems with measurable outputs can indeed have legitimate use, but the misappropriation of the term AI to mean generated content has poisoned the well

=> More informations about this toot | More toots from mgleadow@mastodon.green

Written by Will Deakin on 2024-11-18 at 21:36

@mgleadow ever was it so. If you follow the history of "AI" it has a long history of being hyped snake-oil.

=> More informations about this toot | More toots from wnd@fosstodon.org

Written by Will Deakin on 2024-11-18 at 21:36

@mgleadow maybe we need a better set of words?

=> More informations about this toot | More toots from wnd@fosstodon.org

Written by Martin Gleadow on 2024-11-18 at 21:50

@wnd I was resistant to the co-opting of AI as a cynical marketing tool but it achieved critical mass so I try instead to be more precise about what I mean if I use it correctly

Which is very similar to when "the cloud" was slapped on everything so I talked about *aaS

=> More informations about this toot | More toots from mgleadow@mastodon.green

Written by Will Deakin on 2024-11-18 at 22:18

@mgleadow were still in the XaaS phase. It's just painful.

The problem with the AI schtick is that we're arguably at the third or fourth iteration of its super hype cycle. Each one different, each one inevitably under delivering but equally eventually finding some kind of less hyperbolic level. The then not be AI anymore.

O tempora! O mores!

=> More informations about this toot | More toots from wnd@fosstodon.org

Written by Will Deakin on 2024-11-18 at 21:30

@mgleadow I don't know if this is using AI without a sound, validated and approved use case. It probably isn't. In my case, it's noodling with some algorithms and seeing what it gets me*.

Equally I am not sure if it is a reckless use of finite resources. I pay for the electricity and it runs on a five year old laptop.

• Not very far if truth be told.

2/n

=> More informations about this toot | More toots from wnd@fosstodon.org

Written by Martin Gleadow on 2024-11-18 at 21:33

@wnd I may have been being deliberately dramatically dictatorial ;)

Noodling around with algorithms is not particularly costly. Pumping the "big data" into "the cloud" and running it until something almost but not entirely unlike a cup of tea emerges is

=> More informations about this toot | More toots from mgleadow@mastodon.green

Written by Will Deakin on 2024-11-18 at 21:34

@mgleadow where I completely agree is the industrial abuse of AI at data-centre scale. But I would also add bitcoin mining, spam-email and other spurious abuse of technology.

Equally I'd not want to fall out about it.

3/n

=> More informations about this toot | More toots from wnd@fosstodon.org