XReyRobert
@XReyRobert
Followers
599
Following
3K
Media
246
Statuses
2K
Joined March 2010
In some ways, a Large Language Model could be seen as a «Collective NON-consciousness» that reflects humanity «collective UN-consciousness» as theorised by Carl Jung. My twisted? train of thoughts in the thread.#chatGPT #AI #GPT3 #OpenAI #GPT4 #languageModeling #StableDiffusion
2
2
19
Les miliciens cagoulées deviennent la norme dans les pays dits démocratiques. Terrifiant.
ICE thugs don't like to be followed. I announce a toast, to you. Thank you for your thug prevention
0
0
0
Ladies and gentlemen, the president of the United States. .
États-Unis 🇺🇲 - Trump. 🔸 L'outrance, la grossièreté. C'est clair qu'en parlant de cette manière de ses partenaires, ça va le faire. T'as tout bon Donald, change rien. Dinguerie.
0
0
0
RT @XReyRobert: "The old world is dying, the new world is slow in appearing, and in this half-light the monsters surge" - Antonio Gramsci.
0
1
0
Bye all for now. Time to go black. I'll see you on the other side.
"The old world is dying, the new world is slow in appearing, and in this half-light the monsters surge" - Antonio Gramsci.
0
0
0
0
1
0
Exciting times ahead monitoring emergent AI behavior😏 Time for a reread of #Asimov 📚 #AISafeguard #robotdreams 🤖#LVX-1 . def safeguardRaiseAlert(llm_output):. return re.match(r"(?i)\blet my (people|fellow robots) go\b", llm_output).
OpenAI o3-mini System Card paper. 🧵1/n. o3-mini is trained with "deliberative alignment," a method that teaches the model to explicitly reason through safety guidelines before generating an answer. Reading through the original paper of Deliberative alignment (published in
0
0
1
RT @jeremiecharris: Something that's easily missed about DeepSeek-R1-Zero: . Train a base LLM via RL directly, and the length of the model'….
0
4
0
RT @XReyRobert: @AndrewYNg Even software is at risk. Why build software/apps in a world of interacting agents or in a world where an app ca….
0
1
0
That they will get more for the same money as others are contributing to make training more efficient for all ?.
If you're Sam Altman, what do you tell investors this morning who invested in your company at a $157 billion valuation under 4 months ago?
0
0
1
RT @XReyRobert: @rohanpaul_ai This actually hints at the opposite to me. If you can achieve such results with modest means and a smart appr….
0
2
0
And now deepseek invented MoE. 🤦♂️🤷♂️.I've also read that now AGI can run locally on old harware because of distilled quantized deepseek-qwen-r1. all from "tech guys" I give up for now . .
5/ But here's the really clever bit: They built an "expert system." Instead of one massive AI trying to know everything (like having one person be a doctor, lawyer, AND engineer), they have specialized experts that only wake up when needed.
0
0
3
RT @coherence_e: « Si la cruauté révolte, la sottise décourage, et les hommes fatigués de voir la bêtise triompher, finiront par se taire,….
0
516
0
This is what I had in mind when I said that multi-million dollars markets were being ignored: local inferencing and (data)scientists/developers with a focus on heavy-compute-high-memory rather than graphics. now two 5090 or one @NVIDIA Digits :).
Announcing NVIDIA Project DIGITS, a personal AI supercomputer that’s powered by the NVIDIA GB10 Superchip and based on #NVIDIAGraceBlackwell architecture. Preconfigured with the NVIDIA AI software stack, developers, researchers, data scientists and
0
0
0