Nicola Vitucci | @[email protected]

Jack of all trades, master of some. Created knowledge graphs before it was cool. ♥️ Pythons, OWLs, a SPARQLing life, data, models, robots. Privacy is not dead.

[Nicola Vitucci | @[email protected] @nvitucci]

@kearneymw Friends don't let friends share ?-loaded URLs

[Nicola Vitucci | @[email protected] @nvitucci]

@tunguz Happy Twentycember holidays!

[Nicola Vitucci | @[email protected] @nvitucci]

@Grady_Booch Grady, what do you make of the lawsuit over Copilot?

[Nicola Vitucci | @[email protected] @nvitucci]

You know the elf on the shelf, now get ready for:

[Nicola Vitucci | @[email protected] @nvitucci]

@raistolo @polimi Congrats, Stefano! Absolutely deserved :)

[Nicola Vitucci | @[email protected] @nvitucci]

@tunguz "Train ML models using an open source libra"

[Nicola Vitucci | @[email protected] @nvitucci]

@itsBexli How many times I've been saying this too...

[Nicola Vitucci | @[email protected] @nvitucci]

@danbri Congrats!

[Nicola Vitucci | @[email protected] @nvitucci]

@billyhumblebrag Yes, because GitHub/Microsoft is profiting without giving a lot back, not even recognition (in the shape of attribution from within Copilot, for instance).

[Nicola Vitucci | @[email protected] @nvitucci]

@chrisalbon I am also wondering what the mid- to long-term reaction of open source developers and platforms will be. After all, a class-action lawsuit is already underway.

[Nicola Vitucci | @[email protected] @nvitucci]

Waiting for the time when it's revealed that the output of ChatGPT does not come from the LLM but from a large cache of interactions that 1) people have been filling up and 2) other people have been moderating

[AK @_akhaliq]

Diffusion Art or Digital Forgery? Investigating Data Replication in Diffusion Models
abs: https://t.co/VfVDCq1cwY

[Nicola Vitucci | @[email protected] @nvitucci]

@aazadmmn My point is just that there is an example on Wikipedia which is almost exactly the same, so knowing that the model has been trained on Wikipedia makes this result less surprising.

[Nicola Vitucci | @[email protected] @nvitucci]

@aazadmmn Not sure, given that https://t.co/kUq1znShPC exists

[Nicola Vitucci | @[email protected] @nvitucci]

@DanielMiessler Now I am curious.

[Nicola Vitucci | @[email protected] @nvitucci]

@bentossell https://t.co/S6Y9XxSRHj

[Nicola Vitucci | @[email protected] @nvitucci]

What if GPT-4 is not trained on existing text but just on all the interactions with ChatGPT... (almost) for free

[Nicola Vitucci | @[email protected] @nvitucci]

What if GPT-4 is not trained on existing text but just on all the interactions with ChatGPT... (almost) for free

[Nicola Vitucci | @[email protected] @nvitucci]

@Plinz "Don't tell me how to do humble! My humble ranks among the greatest in the history of the universe!" (Futurama ♥️)

[Davide Eynard @aittalam]

"#PicoGopher Part 6: here comes the sun", or where I harness the power of a flaming star to server gopherholes from a thumb-sized server.

https://t.co/MUsJ7WR3Nm

#gopher #SmallInternet #raspberryPi #raspberryPiPico #SelfHosting

[Nicola Vitucci | @[email protected] @nvitucci]

@fchollet I would also expect that, at that point, skilled people wouldn't want to make their work available for a model to pick it up for free.