AIwithoutFear @JobstLandgrebe

Jobst Landgrebe is a tech entrepreneur, AI researcher and philosopher. https://www.aiwithoutfear.net Joined May 2020

Tweets

190
Followers

326
Following

27
Likes

0

AIwithoutFear @JobstLandgrebe

2 weeks ago

LLM are one of the biggest tech bubbles ever, really a fraud scheme targeting those who confound marketing language with reality. wheresyoured.at/ai-bubble-2027/

0 1 3 78 0

M. Buber was a famous Jewish philosopher who developed a theory of "I and Thou" in the 1920s. Proudfoot believes that such a relation could form between man and machines (running AI: zygonjournal.org/article/id/175…). This is of course impossible since machines will never be conscious.

1 1 2 99 0

AIwithoutFear @JobstLandgrebe

2 months ago

The sobering reality of robo football, two key findings: No active perception, no real intersubjective collaboration. Both impossible to engineer. youtu.be/z5us32WsqJk

1 0 3 301 0

AIwithoutFear @JobstLandgrebe

3 months ago

Darpa wants to model and predict human collective behaviour. This is an old dream of ruling classes for which the Romans used the haruspex. What Darpa plans is of similar effectiveness since mathematical models will miserably fail at that task. thedebrief.org/darpa-wants-to…

0 2 4 261 0

AIwithoutFear @JobstLandgrebe

4 months ago

We won't have freely moving humanoid robots since we cannot configure the reaction to an open world environment. Watch this one to appreciate the state of the art: zerohedge.com/technology/wat…

0 2 1 367 1

AIwithoutFear @JobstLandgrebe

5 months ago

My co-author Barry Smith made a video describing the second edition of our AI book describing the absolute limits of AI: youtu.be/QY-wOcFdA_o

3 2 6 512 2

AIwithoutFear @JobstLandgrebe

8 months ago

New interview with me on AI and transhumanism: youtu.be/0qNlp5Hf5dU with some new aspects.

2 1 5 711 0

AIwithoutFear @JobstLandgrebe

9 months ago

o3 GPT can do well on all parts of IQ tests which resemble its training distributions. It can never do anything that falls outside it. Yet intelligence is the ability to deal with novel situations sponateneously and adequately without prior training. see x.com/garymarcus?lan…

0 0 5 482 2

AIwithoutFear @JobstLandgrebe

9 months ago

Gary Marcus sees clearly that ML (Sora) cannot learn physics as humans can. The reason is, of course, that the human mind is not a clean slate when we are born, but that we have innate mechanisms to process the world which enable our survival. We cannot endow these to machines.

1 1 5 403 0

AIwithoutFear @JobstLandgrebe

10 months ago

My friend and co-authors Barry Smith talking about AI and Virtual reality using material from the upcoming second edition of our book "Why machines will never rule the world". youtube.com/watch?v=SJbXt0…

0 1 2 385 1

AIwithoutFear @JobstLandgrebe

10 months ago

The more insightful AI research colleagues believe that AI will not learn how to conduct dialogues because of too little training material: arxiv.org/abs/2410.16044 They do not realise that language production is non-ergodic, so that training data amounts do not matter.

1 0 1 232 2

AIwithoutFear @JobstLandgrebe

a year ago

While this is to a large extent correct: ojoyoshidareport.com/why-our-machin… to get more details and a more comprehensive argument see our book "Why machines will never rule the world" (2n ed. in prep)

0 2 4 347 3

AIwithoutFear @JobstLandgrebe

a year ago

Our book on the fundamental impossibility of AI is getting more traction: lawliberty.org/book-review/th…

0 1 5 389 2

AIwithoutFear @JobstLandgrebe

a year ago

Some use the Church Turing Deutsch Thesis ("a universal computing device can simulate every physical process.") to support the possibility of AI. But quantum physics fails at many aspects of reality. Deutsch overestimates physics.

1 0 5 529 1

AIwithoutFear @JobstLandgrebe

a year ago

A reader of our book who does not understand the relationship of the human mind and the world: snodgrass.blog/why-machines-w…

2 0 4 520 2

AIwithoutFear @JobstLandgrebe

2 years ago

arcinstitute.org/news/blog/evo Evolution has partial ergodic patterns, and they can be configured into the models. Nucleic acids in isolation are inanimate, and they have regular patterns. It is a different matter altogether when energy flows through a living system.

0 0 3 393 0

AIwithoutFear @JobstLandgrebe

2 years ago

LLM do not understand anything (as we know), here is an example from bioinformatics. See also our book "Why machines will never rule the world" sciencedirect.com/science/articl…

1 1 5 528 1

AIwithoutFear @JobstLandgrebe

2 years ago

The AI hysteria is even driving mathematicians to worry about AI consciousness and ask for tests. We have shown it can’t be engineered or evolve in Why machines will never rule the world. nature.com/articles/d4158…