A worthwhile article by Gary Marcus for people interested in LLMs and the limitations of LLMs.
Open the article and search for 'Hanoi'.
The well-known Tower of Hanoi is a puzzle that can be solved by junior programmers, but above 7 floors (disks), LLMs simply crumble under the complexity.
As Marcus says: "AI is not hitting a wall. But LLMs probably are."
https://garymarcus.substack.com/p/a-knockout-blow-for-llms
#LLM #AI #Puzzle #TowerOfHanoi #GaryMarcus
Gary Marcus: A knockout blow for LLMs?. “There’s actually an interesting weakness in the new argument—which I will get to below—but the overall force of the argument is undeniably powerful. So much so that LLM advocates are already partly conceding the blow while hinting at, or at least hoping for, happier futures ahead.”
https://rbfirehose.com/2025/06/10/gary-marcus-a-knockout-blow-for-llms/
Two more reasons why every politician or organisation should leave X if they didn't already,
Elon Musk's vison on AI, Grok3 is integrated in X.
https://garymarcus.substack.com/p/elon-musks-terrifying-vision-for?triedRedirect=true
Also, X blocking Signal,
https://www.disruptionist.com/p/elon-musks-x-blocks-links-to-signal
Sam Altman says “we are now confident we know how to build AGI” - On Sunday, OpenAI CEO Sam Altman offered two eye-catching predictions abou... - https://arstechnica.com/information-technology/2025/01/sam-altman-says-we-are-now-confident-we-know-how-to-build-agi/ #simulatedreasoning #superintelligence #claude3.5sonnet #machinelearning #garymarcus #anthropic #samaltman #srmodels #biz #claude #o1-pro #openai #agi #ai #o1 #o3 #sr
Ofcourse @garymarcus was right when he predicted that the top foundation models would encounter diminishing returns and could be in some ways in a dead-end street on the road to "AGI". However current models with current capabilities will still cause a big revolution/disruption. NotebookLM is a good example, that works on basis of Gemini 1.5, not the best model, still the implementation/application made the difference.
#AI #garymarcus #LLM #GoogleGemini #NotebookLM #OpenAI
Good sceptical article by Gary Marcus, regardless the fundamental problems with current LLMs , progress is remarkable however in my opinion. So reliable agents with LLMs are a problem, many other applications are not,
https://garymarcus.substack.com/p/llms-dont-do-formal-reasoning-and
#ai #garymarcus
#LLM don’t do formal reasoning - and that is a HUGE problem #AI #GaryMarcus
Important new study from #Apple
https://garymarcus.substack.com/p/llms-dont-do-formal-reasoning-and
In 2023, #AI firms collectively spent $50 billion on #Nvidia chips alone. They brought in only $3 billion in revenue.
AI researcher and cognitive scientist #GaryMarcus has said that the industry is dealing with a simple, core problem: #hallucinations make models unreliable; that unreliability makes them non-viable for #corporate use-cases, #AIBubble #GenerativeAI
Here's how investors should navigate the bubble of the #AIBoom
https://www.thestreet.com/technology/heres-how-investors-should-navigate-the-bubble-of-the-ai-boom
Gary Marcus again ... writing about the threat of information pollution
"... the problem is quickly getting worse."
"If we don’t get a handle on this problem, fast, nobody will believe anything."
Overly pessimistic or just (beginning) reality?
https://garymarcus.substack.com/p/information-pollution-reaches-new
Gary Marcus peer reviewed an Open AI paper, interesting read.
#ai #generativeAI #gpt4 #openai #garymarcus
https://open.substack.com/pub/garymarcus/p/when-looked-at-carefully-openais
According to AI expert Gary Marcus, generative AI - the technology that can create realistic content from text, images, or audio - is facing a swift end. Marcus argues that it is based on flawed assumptions and lacks common sense, creativity, and generalization. He warns that it is a dead end that will not lead to human-like intelligence or solve real-world problems.
Wanted to hear the interplay and in situ there is good surmise of current status salience, but if you zoom out also a bit of a localized fart bubble because in a global sense we have a whole cluster of more immediate threats on the scale of doom than #AI - problems that right now can't be kicked down the road without blowing up in our faces on a much shorter timescale
Will #AI Destroy Us ? Roundtable discussion with #ScottAaronson , #EliezerYudkowsky and #GaryMarcus
".... And, bonus, there is little if any government regulation in place to do much about this. The possibilities are now endless for propaganda, troll farms, and rings of fake websites that degrade trust across the internet." #ChatGPT #GaryMarcus
Why everyone should be at least a little bit worried about AI.
What have dozen of scientists, economists, researchers, and elected officials all have in common?
They are all worried about the near-term future of AI. The most worrisome thing of all? They are all worried about different things.
https://garymarcus.substack.com/p/an-epic-ai-debateand-why-everyone
#AI #ArtificialIntelligence #dangers #neuroscience #neuroscientis #social #socialresearch #research #causal #causality #GaryMarcus @garymarcus
#sociology
On the other side is a discussion in a recent podcast episode of Ezra Klein's podcast titled "A Skeptical Take on the AI Revolution" where he and "AI expert" Gary Marcus discuss these accuracy and reliability challenges in more detail. I'm still listening to the podcast episode so I haven't heard the whole discussion just yet.
You can find the episode on Pocket Casts here: https://pca.st/m2vc0n12
You can also find the podcast behind the New York Times paywall here: https://www.nytimes.com/2023/01/06/opinion/ezra-klein-podcast-gary-marcus.html
How are the activities & assessments we design as teachers for our students “different than an auto-complete activity?”Great conversation with about #OpenAI #ChatGPT with #EzraKlein & #GaryMarcus
«Este veterano de la inteligencia artificial explica por qué ChatGPT es "peligrosamente estúpido"»
El estadounidense Gary Marcus (52 años), profesor emérito de la Universidad de Nueva York, es uno de los popes de la IA tan conocido por sus aportaciones (fundó dos empresas de IA, compradas por Uber; ha escrito decenas de papers y 5 libros, el último, Rebooting AI , considerado una de las obras clave)
"𝚂𝚒 𝙲𝚑𝚊𝚝𝙶𝙿𝚃 𝚕𝚕𝚎𝚟𝚊 𝚊 𝚞𝚗𝚊 𝚙𝚎𝚛𝚜𝚘𝚗𝚊 𝚊 𝚌𝚛𝚎𝚎𝚛 𝚚𝚞𝚎 𝚎𝚜 𝚒𝚗𝚝𝚎𝚕𝚒𝚐𝚎𝚗𝚝𝚎 𝚌𝚞𝚊𝚗𝚍𝚘 𝚗𝚘 𝚕𝚘 𝚎𝚜, 𝚎𝚜𝚎 𝚎𝚜 𝚎𝚕 𝚙𝚎𝚕𝚒𝚐𝚛𝚘"
Marcus sostiene que ChatGPT no es el camino a seguir para lograr máquinas inteligentes. "Es divertido, te entretienes un rato, pero es como poner a un mono delante de un teclado".
"𝙴𝚕 𝚌𝚑𝚒𝚜𝚝𝚎 𝚍𝚎𝚕 𝚋𝚘𝚛𝚛𝚊𝚌𝚑𝚘 𝚋𝚞𝚜𝚌𝚊𝚗𝚍𝚘 𝚕𝚊𝚜 𝚕𝚕𝚊𝚟𝚎𝚜 𝚋𝚊𝚓𝚘 𝚕𝚊 𝚕𝚞𝚣 𝚍𝚎 𝚞𝚗𝚊 𝚏𝚊𝚛𝚘𝚕𝚊 𝚍𝚎𝚜𝚌𝚛𝚒𝚋𝚎 𝚖𝚞𝚢 𝚋𝚒𝚎𝚗 𝚚𝚞é 𝚙𝚊𝚜𝚊 𝚑𝚘𝚢 𝚌𝚘𝚗 𝚕𝚊 𝙸𝙰"
► @garymarcus ◄
