Marcus M

[Marcus M @MarcusMLA1]

@blader That means Democrats win 🥇

[Marcus M @MarcusMLA1]

@scaling01 They are just going to extend until Opus 5 is released, then Fable gets moved to API

[Marcus M @MarcusMLA1]

@ChrissGPT I had the exact same experience a few months ago when I picked up my model y and was told to use FSD when driving home so that they can make sure everything is calibrated correctly. The car drive me home and it was the most surreal experience of my life. Tesla for life!

[Marcus M @MarcusMLA1]

@DeryaTR_ I don’t think a slowdown could be achieved even if they wanted to because of China/USA competition and other forces. What I think may end up happening is nationalization and possible merger of AI companies. The USG wants to accelerate AI development but restrict public models.

[Marcus M @MarcusMLA1]

@VraserX @Exuma72 Good luck with that 😆

[Marcus M @MarcusMLA1]

@VraserX Anyone who expected anything else but tiered and gated releases as capability thresholds cross is just naive.

[Marcus M @MarcusMLA1]

@kimmonismus Don’t you think that the developers of open source models will stop giving out their frontier models for free? Thus far a lot of open source models have been distilled but that could stop rapidly; developing a frontier model costs billions potentially…open source may be cooked

[Marcus M @MarcusMLA1]

@DeryaTR_ Yes it’s all about profit maximizing…also cyber security is a big revenue potential so expect major restriction and capability downgrades along with all bio, chem and some other science.

[Marcus M @MarcusMLA1]

@ChrissGPT @nicdunz So true

[Marcus M @MarcusMLA1]

@VraserX That will happen regardless of which party is in power or gets elected

[Marcus M @MarcusMLA1]

@PrometheanPrt @DaveShapi You are surprised capitalism is cut throat? Ever worked a corporate job in America? What did you expect would happen?

[Marcus M @MarcusMLA1]

@tunguz That actually sounds like the most likely path tbh…

[Marcus M @MarcusMLA1]

@beffjezos It’s not going to slow down innovation…ai labs will still accelerate…you can’t have a public model capable of rsi unless you somehow control the user and usage. There are lots of companies and scientist who will accelerate science and progress…but public will not get agi

[Marcus M @MarcusMLA1]

@VraserX This was predicted long time ago or at least around 2024 by Leopold Aschenbrenner in situational awareness…Superintelligence will be kept guarded by default…but that’s ok it doesn’t mean progress will stall, the opposite, ai will super charge science, physics etc.

[Marcus M @MarcusMLA1]

@haider1 That’s not true…frontier models are being restricted and guardrails put on because their capabilities crossed certain thresholds and have risks…this was always expected and predicted by many ai insiders including Leopold Aschenbrenner and many others…new era has begun

[Marcus M @MarcusMLA1]

@AndrewCurran_ This doesn’t meant that ai progress will slow down, actually I think progress is going to accelerate…but the strong models will be restricted to partner and govt and not released as public products. Public will get the second hand benefits of the ai acceleration.

[Marcus M @MarcusMLA1]

@DeryaTR_ @OpenAI Congratulations

[Marcus M @MarcusMLA1]

@haider1 AI development is not pausing, quite contrary its hyper accelerating soon via RSI which ironically is the exact reason why public releases are being restricted and guarded. Acceleration will continue with ever increasing dichotomy between internal and external releases.

[Marcus M @MarcusMLA1]

@hadesboun101 @VictorTaelin Not really…there will be deployment, API, import/export restrictions and other means including global ai deployment deals that will nerf open source models also…already USG is putting pressure on META models including open source.

[Marcus M @MarcusMLA1]

@rand_longevity @VictorTaelin This will be the new normal by necessity…powerful models will be kept internal and or sandboxed while public will get consumer and productivity tools/toys…medical and other breakthroughs will still accelerate, just not via public models