I'm an inventor, scientist and engineer. My passion is in modeling the brains of animals to make machines more useful - with language. We're getting there!
Palo Alto, CA
Joined June 2010
- Tweets1.5K
- Following200
- Followers560
- Likes2.3K
John Ball retweeted
Not entirely correct @ChombaBupe . Cognitive scientist @jbthinking has a working implementation for Language understanding on a machine from his brain theory.
We agree today’s generative AI has more to do with dangling carrots than brain emulation.
John Ball retweeted
Me: How does one solve spatial intelligence?
Fei-Fei Li: By creating 3D world simulations.
Me: I see. Well, good luck with that. 🙄
PS. It's a good thing babies don't need a 3D computer simulation to gain spatial intelligence, isn't it? The real world is free and perfect. 😀
John Ball retweeted
How do you achieve the brain’s language and vision comprehension? You model the brain using empirical science then implement on a machine as a means for a continuous feedback loop.
@jbthinking writes about that implementation for language in ‘How to solve AI with Our Brain’
John Ball retweeted
@jbthinking ‘s book ‘How to Solve AI with Our Brain’ - Pat 2 talks about the brain model required for robotics.
John Ball retweeted
Prediction
No Optimus robot, regardless of how advanced the engineering is, will ever be deployed in the home (or build human habitats on Mars) without someone cracking intelligence first. If the Optimus team's approach is to collect training data for the robot's brain, they already failed imo.
Babies grow up in the real world and become CEOs, programmers and roboticists without ever collecting training data. How do they do it? They use their sensors and effectors to learn continually, interact with and achieve goals in the real world.
Real world robotics means deployment and learning in the real world. Everything must be learned from scratch, including grabbing stuff, crawling, walking, running, talking, etc. 🫤🤔
"Elon Musk is just an extraordinary engineer, and I love working with him. We're going to build many more computers together. I'm delighted to be working with him. I think the Optimus opportunity is just right around the corner. It's very likely that human robots are going to be robots that we can deploy into the world relatively easily & this is the first robot that really has a chance to achieve the high volume and technology scale necessary to advance technology and I think this is likely to be the next multi trillion dollar industry."
一 Nvidia CEO Jensen Huang
John Ball retweeted
35 million people laughed in my face.
But we still don’t have a solution to hallucinations, boneheaded errors, and unreliable reasoning.
Years later, the wall of reliability still looms in front of LLMs, unconquered.
John Ball retweeted
The brain starts learning from the moment you're born. Do you want to experience that hyper growth you did as an infant?
Come with us on the #SpeechGenie journey.
#languagelearning
John Ball retweeted
OpenAI has no choice but to triple down on sunk costs including more marketing costs. The architecture is not the path to 'AI' and the flawed model is showing up as a game of 'wack-a-mole' and constant apologies.
The next generation with the fit for purpose model is the cognitive science lead one. @jbthinking writes about it here:
a.co/d/cKkJcJ6
John Ball retweeted
Like hasn't been achieved for ~70 years the promise of 'AI' or 'AGI' is plagued with hype or as some call it, 'hopium'. The more costly effort in both💰 and 🌍resources committed to chase the holy grail the further the industry goes away from the trustworthy and energy efficient solution.
What solution is that? The that emulates language and vision comprehension of the brain. @jbthinking writes about the cognitive breakthroughs that have occurred to get us to that accurate and energy efficient by copying the old gold standard - 🧠. You can read more in his book 'How to solve AI with Our Brain' and what it takes to benefit the whole of society.
a.co/d/9y1XcQt
John Ball retweeted
Please. The AI community has made zero progress in solving AGI since the foundation of the field in the 1950s. Zilch. They've only worked on computer automation while pretending to be researching intelligence.
Progress is incremental in engineering. In science, progress goes from one breakthrough to the next.
John Ball retweeted
Creating a Better Benchmark by @jbthinking medium.com/pat-inc/creating-…
John Ball retweeted
Many experts in the field feel like you @ChombaBupe . In a AAAI survey 75% of experts believed today's AI was *not* on the right path.
It's those building or researching how to make the best tools for society that resist the hype that goes with today's generative AI. As you point out with many others the problems and collateral damage outweigh any benefits.
More consumption of dollars and computer resources will only produce more BS spewing pollution requiring a constant whac-a-mole for enterprise developers who require accuracy for language and vision comprehension. For that, there is only one gold standard - the brain. @jbthinking writes about that next generation of tools that 'understand' as well as 'generate':
a.co/d/3gRROdE
John Ball retweeted
Learn like a child and speak like a native. @kungfu_english dispels myths while showing the brain-friendly learning methodology for adult language learners.
'Something that makes you understood, makes you understand'
a.co/d/g41Shrc
John Ball retweeted
@jbthinking writes about the '7 killer apps' that are currently waiting for that next generation 'AI algorithm'. Like Google wasn't the first search algorithm, generative AI for language and vision will be succeeded by the cognitive science approach. Because today's whack-a-mole experience for software developers of autonomous driving and digital agent applications won't cease until that 'general algorithm for driving' arrives.
For that cognitive science path that empowers society with control over communications with machines, follow John and read his book. Link in comments below
👇👇👇👇👇👇👇👇👇👇👇👇👇👇
𝗔𝘀 𝗜 𝘄𝗮𝗿𝗻𝗲𝗱 𝗶𝗻 𝟮𝟬𝟭𝟲, 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝘁𝗿𝘂𝗹𝘆 𝗿𝗲𝗹𝗶𝗮𝗯𝗹𝗲 𝗮𝘂𝘁𝗼𝗻𝗼𝗺𝗼𝘂𝘀 𝘃𝗲𝗵𝗶𝗰𝗹𝗲𝘀 𝘄𝗶𝘁𝗵 𝘁𝗵𝗲 𝗳𝗹𝗲𝘅𝗶𝗯𝗶𝗹𝗶𝘁𝘆 𝗼𝗳 𝗵𝘂𝗺𝗮𝗻 𝗱𝗿𝗶𝘃𝗲𝗿𝘀 𝘄𝗼𝘂𝗹𝗱 𝗽𝗿𝗼𝘃𝗲 𝗲𝘅𝘁𝗿𝗲𝗺𝗲𝗹𝘆 𝗱𝗶𝗳𝗳𝗶𝗰𝘂𝗹𝘁. 𝗜𝘁 𝗵𝗮𝘀. 𝗛𝗲𝗿𝗲’𝘀 𝗮 𝘀𝗮𝗻𝗶𝘁𝘆 𝗰𝗵𝗲𝗰𝗸:
In 2019, Elon Musk promised there would be a million robotaxis on the road by the end of 2020. So far he has none, and will have a handful (a few hundred?) next year.
Waymo, further along, just announced that it is about to roughly double its fleet, but that still gets them only to about 3,500 across four (soon seven) cities, still less than 1% of the 1,500 or so cities in the US, and less than 0.01% of the 150k cities in the world. Bringing a new city online remains a major investment of time and money.
Furthermore, driverless cars, over a decade after Sergey Brin claimed they would be routine by 2017, still require teams of remote operators, with unknown number of employees, wizards behind curtains. (At one point, Cruise more people in remote operations than they had cars on the road).
The overall investment in driverless cars has been greater than $100 billion. Net revenue no more than a couple billion. Nobody has built a fully general algorithm for driving despite immense amounts of natural and synthetic data.
The lessons for GenAI hype should be pretty obvious.
John Ball retweeted
The scraping and stealing of creators' IP for training data doesn't need to occur for 'AI/AGI' to provide benefit for society.
While the tech giants & 'AI CEOs' distract with marketing hype and false assumptions of what their generative AI technology can do, the 'E=MC^2' solution is ignored.
The path to that beneficial, trustworthy human-machine interface is paved with cognitive science, *not* more and endless training data.
By supporting @jbthinking 's approach so we can receive attention from investors society will see the 'flip phone to smart phone' speed of adoption. We all want to interact with devices to control them in our own language and we can with the right brain-like approach.
a.co/d/75neSYi
John Ball retweeted
@jbthinking talks about the machine implementation of language understanding and generation from his brain model in his book 'How to solve AI with Our Brain'.
Unlike LLMs, it shows the path forward for a lossless architecture to store human-like context. That is then accessible with the same architecture for the human-like language interface
a.co/d/dW9hzCU
John Ball retweeted
Breakthrough in cognitive science will disrupt marketing budgets. But when?
As 'announcements' and 'breakthroughs' intensify by the tech giants there has been more push back and concern this week than I've ever seen on social media.
No-one can compete with the coercion of the tech giants fueled by their marketing budgets. But 100 years since Kellogg showed the value of advertising, pales in significance compared to hundreds of millions of years of human and animal evolution. Why does that matter? Because marketing doesn't change the human brain's capabilities and hence the complexity of mimicking it for machines.
When @jbthinking wrote his book based on brain science, he included the sub-title 'The final Frontier in Science' because science wins every time over enormous marketing and scaling budgets.
You can read about it now in 'How to Solve AI with Our Brain - the Final Frontier in science' to be ahead of the pack. Chapters 18-20 articulate the future including John's 5 year vision when this architecture is given investment required to enable at least '7 killer apps' (Chapter 18, Enabling AI with Science).
a.co/d/dW9hzCU
John Ball retweeted
Let's build dictionaries for the world's languages *once* with an architecture that enables learning on the fly.
How? You can read more about it in @jbthinking 's new book 'How to solve AI with Our Brain'
John Ball retweeted
@jbthinking has described todays AI as in a scientific crisis. Desperately scaling with ‘hopium’.
Clever engineering alone won’t get to human like language and vision comprehension. That requires a scientific breakthrough