@ids@c.im
@ids@c.im avatar

ids

@ids@c.im

Happiest when making stuff or making people laugh. Tech, and Design Thinking. Since 2019, freelancer and FRSA. Formerly https://mastodon.social/@ids on mastodon.social and https://twitter.com/ids

This profile is from a federated server and may be incomplete. Browse more on the original instance.

ids, to random
@ids@c.im avatar

Having an excellent day so far at Thinking Digital - origami and engineering #TDC24

ids,
@ids@c.im avatar

@coldclimate next year!

matthewskelton, to random
@matthewskelton@mastodon.social avatar

Current earworm 👂🎵 - Ilkley Moor baht 'at ... in a swing / Big Band style 🤷🤦

https://en.wikipedia.org/wiki/On_Ilkla_Moor_Baht_%27at

ids,
@ids@c.im avatar

@matthewskelton is it this? https://youtu.be/FFWzt6xYE2w with love from Ilkley :)

ids,
@ids@c.im avatar

@matthewskelton I secretly rather like that version

matthewskelton, to ai
@matthewskelton@mastodon.social avatar

🔮 Prediction: "small" language models with what we might call a "micro corpus" of text will be more useful to most organisations than a general purpose LLM like ChatGPT.

GenAI models trained on a small, focused body of materials so they are highly domain specific, all provided as a service, with the model private to each organisation.

#AI #ChatGPT #LLM

ids,
@ids@c.im avatar

@matthewskelton @ghrasko I think the point is that it’s more helpful to see the LLMs as reasoning engines rather than knowledge repositories.

The architecture for making an LLM into a domain expert is to provide it with the documents and information that #you want it to be an expert on as embeddings in a vector database such as pinecone. The information can then be searched for the parts that are most semantically relevant to a particular user query, and then the retrieved info can then be supplied to the LLM as context along with the user query.

I found this short course particularly illuminating on this topic -> https://www.deeplearning.ai/short-courses/langchain-chat-with-your-data/

ids,
@ids@c.im avatar

@matthewskelton @ghrasko Yes, 100% agree on the eco aspects. The side loading also makes it less likely that the LLM will be making stuff up that it doesn’t know as there’s much less reliance on its “knowledge”.

christianselig, to random
@christianselig@mastodon.social avatar

Judging by the amount of Apollo merch that’s sold in the last two weeks without me even mentioning it, maybe I should be in the t-shirt biz instead of apps

So, if you want to pick some up, there’s a bunch of stuff available at https://cottonbureau.com/people/apollo

ids,
@ids@c.im avatar

@christianselig perhaps you should include a warning for Reddit employees not to wear these in the office

matthewskelton, to ChatGPT
@matthewskelton@mastodon.social avatar

LLMs and GPT-based AI systems should pay IP fees to all creators and authors whose work provides the AI ability.

What's that you say? They wouldn't be financially viable? Then these AI systems are basically stealing. 💰

ids,
@ids@c.im avatar

@matthewskelton I think it’s quite likely that due to the huge numbers involved the contribution of value from any one artist (or the amount by which the LLM would be less valuable without it) would be so tiny as to be pointless.

Surely trying to do this would be a massive amount of admin and cost to almost no benefit?

Perhaps there are other ways the companies could recognise the contributions of all those creators - sponsoring young writers and artists or something similar?

ids,
@ids@c.im avatar

@matthewskelton Now steady on... you can't just make large tech companies go around paying tax. It'd be the end of civilisation as we know it! Or something.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • megavids
  • thenastyranch
  • rosin
  • GTA5RPClips
  • osvaldo12
  • love
  • Youngstown
  • slotface
  • khanakhh
  • everett
  • kavyap
  • mdbf
  • DreamBathrooms
  • ngwrru68w68
  • provamag3
  • magazineikmin
  • InstantRegret
  • normalnudes
  • tacticalgear
  • cubers
  • ethstaker
  • modclub
  • cisconetworking
  • Durango
  • anitta
  • Leos
  • tester
  • JUstTest
  • All magazines
    •