Peter Voss
@peterevoss
Followers
2K
Following
16K
Media
196
Statuses
2K
Creating AGI: https://t.co/1vZB37ewBr https://t.co/EMZMhH4i2n https://t.co/O1HZFZOUV9
Austin, TX
Joined May 2009
A new article explaining the difference between Statistical and Cognitive AI -- like chalk and cheese, or like trying to get to the moon via longer ladders instead of rockets... (link in first comment)
5
4
16
Yes, a big problem
Biggest issues in my opinion are: - not showing your tweets to your followers - link deprioritization - promoting low-effort slop over information-dense content
0
0
1
from @Srini_Pa The industry treated scale-first like a religion, itās time to grow up and learn-on-the-fly. This weekās TechCrunch piece finally said the quiet part out loud: the scaling race is running on diminishing returns. Cohereās former VP of AI Research is openly
2
0
5
Why We Donāt Have AGI Yet Abstract ā TheoriginalvisionofAIwasre-articulatedin2002viathetermāArtificialGeneral Intelligenceā orAGI.ThisvisionistobuildāThinkingMachinesāā computersystemsthatcanlearn,reason,andsolve
2
1
6
Because LLMs are brute-force statistical systems that cannot adapt (learn) on the fly. What we need is Cognitive AI
GenAI is exactly the mess I warned you it would turn out to be:
2
1
8
well bummer - you still can't reliably edit your posts... .. a talk I gave at the Singularity Summit in SF in 2007, not 2027!
2
0
1
Grok got the date of my original article wrong, it was actually 2002. This was my contribution to the original Artificial General Intelligence book https://t.co/0W2XtZxL54 And here a link to the talk recoding:
1
1
1
A blast from the past. grog reminded me of a talk I gave at the Singularity Summit in SF in 2027: "Improved Intelligence, Improved Life" Here's Grok's summary: (more in reply) Core Theme: Your talk focused on the transformative potential of Artificial General Intelligence
1
1
2
Yes! Robots will need a brain that can learn in real-time and that has meta-cognition (think about its own knowledge and reasoning) They need Cognitive AI - or what DARAP calls 'The Third (and final) Wave of AI'
Meta's Chief AI Scientist Yann LeCun offers a critical take on the humanoid robot boom. Speaking at MIT, LeCun claimed the "big secret" of the industry is that current companies "have no idea" how to make their robots "smart enough to be generally useful." He argues that while
0
4
11
Two AGI futures to choose from: 1) CONTROL: data extraction, manipulation, surveillance, data weaponization, freedom erased. 2) HUMAN: potential unlocked, dignity, abundance, flourishing, autonomy protected. So, we created this. Will AGI companies pledge? We pledge. š§µ
2
2
2
Here's the pattern we see with investors, when we present the most credible & direct path to AGI via a non-LLM, cognition-first architecture with real-time learning at its core: Almost all investors come to the table assuming intelligence must be bought with massive data, GPUs,
7
4
27
... what are the core principles for creating AGI? We need to understand intelligence in order to build it. The key question to ask is what makes human intelligence to powerful, so special? What is its essence? Insights from epistemology (theory of knowledge) highlight the
Using reinforcement learning for reasoning is a category error.
0
1
7
Real AGI needs the ability to learn new knowledge and skills incrementally in real-time. Whether AGIs are employed as researchers, workers, personal assistants or some other function, a common requirement is that they are able to adapt to changing situations and requirements on
Good to see the market finally noticing the missing piece: Continuous Learning. But these LLM āfixesā are just clever patches for catastrophic forgetting, not systems that learn, evolve, or update a world model in real-time. Most people assume that continuous learning is
1
2
11
LLMs come pretrained with oceans of data from the internet before the deployment. Real Intelligence, AGI, learns via real-world interaction, reflection and autonomy after deployment. It needs a cognition-first architecture with real-time incremental learning at its core.
0
2
6
The utter irony of Putin and Trump meeting soon in Budapest to jointly seal Ukraineās fate should not be lost on anybody. This is the very same city where the US, the UK, and Russia came together 31 years ago to give Ukraine ironclad assurances that its territorial integrity
0
8K
26K
$100 Billion Federal Contracting Scam Exposed: 8(a) Firm Admits to Violating Federal Law, Using Minority-Owned Status as a Front to Obtain $100M+ No-Bid Government Contracts While Outsourcing 80% of the Work. ATI Government Solutions Contract Manager, Melayne Cromwell Admits to
2K
14K
38K
This cycle will repeat for AGI. No AGI without a cognition-first architecture: persistent memory, causal world models with real-time incremental learning at the core. Thatās how you get real intelligence, beyond fluency & some utility, also It shatters the scale-first cult.
Jensen Huang explained why the world laughs at pioneers before calling them geniuses:
1
4
18