The economics of AI stem from there. First study the human brain.
1 Like
It is an interesting article that deserves to be read. Here is a gift link.
3 Likes
Dr. Chomsky is right for the wrong reason, The system is not well designed… it is evolved!
Based on my experience as a computer programmer, thinking about how the brain works, my thoughts were that we have two kinds of brain, the lizard or subconscious brain that controls the fight or flight instinct. There is no time to think when faced with a hungry lion or bear, just pick the best option, fight or flight, and hope like hell it’s the right one. The other brain is the boolean rational brain that does the rational thinking.
To understand the subject of the article one first needs to understand what words are. In nature there are physical objects that we can sense, see, feel, touch, smell, or taste. Humans also have unembodied objects such as love, mercy, mathematics, nationality, right, wrong, and many others for which we create words so that we can communicate them. Words are for communicating. Other animals use scents, songs, colors, and other forms of communication.
Neural networks don’t think in the boolean sense, they just compute probabilities, which is the reason that current large language models can string together words but cannot tell fantasy from reality. That requires boolean thinking, the use of the other brain.
Real or fully functional AI will exist when neural networks can hand off to heuristics when boolean thinking is required. In other words, heuristics fact checking large language model output. It is interesting to note that Richard Feynman stated that science starts with a guess, an educated guess. The guess is then subjected to experiment to separate fact from fantasy. Large language models generate the educated guesses!
It’s interesting that scientists can discover where these brains reside inside our skulls and confirm the existence of brains with diverse functions.
Chomsky is wrong about other things as well but that becomes political.
The Captain
5 Likes
Mammograms aren’t the only type of medical imaging getting AI assistance. Doctors are using the technology to scan X-rays of people’s chests, ultrasound videos of infants’ hearts and more. AI technology in medicine is growing at a rapid clip, and “imaging is leading the way,” said Stanford University radiologist Curtis Langlotz at The New Wave of AI in Healthcare symposium in New York City in May.
… even experts can make mistakes. A radiologist’s daily error rate may be around 3 to 5 percent, researchers have found. Such errors tend to stem from being overworked, scientists reported in 2023 in the European Journal of Radiology. “We need help,” Langlotz said.
2 Likes
It’s not too often that we enter the area of neural networks on this site. I am quite passionate about this subject. The obvious disconnect is, that few see that CNNs (convoluted neural networks) are old. As NVIDIA was able to align everyone to CUDA, they print money with their GPUs.
However, as investors, we want to be a step ahead and should be aware that CNNs will get replaced by SNNs (spiking neural networks).
They mimic natural neural processes more closely than CNNs - by incorporating the timing of neuron spikes into their computation.
This brings, conservatively estimated, a factor 1000x lower energy consumption. Advantages are obvious and need no further elaboration.
Why is it then that this evolution, clearly advertised in Gartners IT wave (Neuromorphic computing), is getting so little attention? Most of us are trend followers. Chances of failure are too high investing too early.
Well, in case someone is interested to extend the shelf life of this discussion… feel free to comment.
[/quote]
6 Likes
So much new to learn!
Investing in AI
A risk averse investor should wait for a technology to Cross the Chasm before investing as many great ideas die at the chasm. SNN technology is nowhere near the chasm while CNN technology clearly has succeeded in this right of passage.
At this time chip makers (NVDA, ARM, AMD) are clear beneficiaries as are the hardware assemblers (DELL, PSTG, SMCI). Looking at a more distant horizon, I would concentrate on companies that will be able to generate useful AI and, more importantly, be able to monetize the AI.
Apple’s iPhone is a good model. Lots of companies make smartphones but Apple’s secret sauce is the Human User Interface (iOS) they started developing for Mac over 40 years ago.
It takes very deep pockets to afford the data centers to train AI models which limits the entrants but that is not enough, you need salable products based on the AI to monetize the AI. How are Google, Facebook, Amazon, Microsoft, et al going to do so?
My current best candidate is Tesla with Intelligent Mobility and Humanoid Robots. BTW, the EVs and the robots will all use the same Tesla developed, ARM based, inference computer. The current version is Hardware 4 (HW-4) and Tesla is already working on version 5 to be called AI-5.
The Captain
3 Likes