27 March 2024

A Brief History of Intelligence (2023) - Max Bennett

Max Bennett has been working in AI for consumer markets, and this book is an ambitious overview that might better be titled 'Breakthroughs in Human Brain Development'.

His history is organized by five evolutionary brain 'breakthroughs' that created/supported new behaviors and capabilities in the development chain along the way toward Homo Sapiens. In brief these are: steering, reinforcing, simulating, mentalizing and speaking. In part he uses this framework to discuss how AI techniques have attempted to create similar capabilities and solve similar problems. For instance in the reinforcing section (esssentially about learning) there are challenges: how does the organism manage to learn something new without losing the last thing it learned, and how to learn when the outcome of an action does not have an immediate result - i.e. overcoming temporal delays.

In the speaking section, he indicates that recent findings indicate not so much that humans have new neocortex areas that support language and speaking, but that somehow we developed base instinctual behaviors that support learning language - proto-conversations with babies, and joint attention - and the other primates don't have these instincts. He describes this capability as the original singularity, since it allows for the growth of knowledge over time that just builds and builds. He thinks that modern man does not have any actual brain capabilities that weren't available tens of thousands of years ago, but we do have the accumulation of knowledge of many generations.

Note that this book does not use the word 'consciousness' at all (as far as I noticed), and the word 'mind' is used sparingly, mostly in the term 'theory of mind' (covered in the mentalizing section). One paragraph stood out to me in particular that indicates a pretty reductionist attitude - this is from page 301:

When we talk of these inner simulations, especially in the context of humans, we tend to imbue them with words like concepts, ideas, thoughts. But all these things are nothing more than renderings in the mammalian neocortical simulation. When you "think" about a past or future event, when you ponder the "concept" of a bird, when you have an "idea" as to how to make a new tool, you are merely exploring the rich three-dimensional simulated world constructed by your neocortex. It is no different, in principle, than a mouse considering which direction to turn in a maze.

While in some sense I agree that these processes are indeed the workings of the brain, I think there's a lot more to be explored here than Bennett describes. Still, I felt this was a worthwhile and interesting book, and I'd like to think about it more in comparison with Kevin Mitchell's 'Free Agents' which covers a similar evolutionary path and history.

More on A Brief History of Intelligence.

20 December 2023

Honest Placebos

I've come across a few things referencing placebos lately, in particular 'transparent' or 'open' placebos where the fact that it contains no known effective ingredient is not hidden.

One is a link to this research on "Effects of open-label placebos in clinical trials: a systematic review and meta-analysis" from Nature dated Feb 16, 2021:

Open-label placebos (OLPs) are placebos without deception in the sense that patients know that they are receiving a placebo. The objective of our study is to systematically review and analyze the effect of OLPs in comparison to no treatment in clinical trials.

We found a significant overall effect (standardized mean difference = 0.72, 95% Cl 0.39–1.05, p < 0.0001, I2 = 76%) of OLP. Thus, OLPs appear to be a promising treatment in different conditions but the respective research is in its infancy.

Then in perusing Andy Clark's latest book The Experience Machine, which posits the brain as a prediction engine, constantly engaging with sensory input both consciously and unconsciously to enable action, he concludes with some material about what he refers to as 'honest' placebos:

Honest placebos appear to work by activating subterranean expectations through superficial indicators of reliability and efficacy such as good packaging and professional presentation (foil and blister packs, familiar font, size and uniformity of the pills, and so on). This is because - as we have seen - the bulk of the brain's prediction empire is nonconscious.

Clark reviews a number of other findings in his 'Hacking the Prediction Machine' chapter, and in a sense concludes:

In the end, it looks like anything that can be done to increase our confidence in an intervention, procedure, or outcome is likely to have real benefits. 

He also describes use of certain psychedelic drugs as having the potential to 'reset' the prediction machine in very useful ways.

18 December 2023

Conversing with a whale

This Dec. 12, 2023 report from the Seti Institute, Whale-SETI: Groundbreaking Encounter with Humpback Whales Reveals Potential for Non-Human Intelligence Communication seems encouraging.

In response to a recorded humpback ‘contact’ call played into the sea via an underwater speaker, a humpback whale named Twain approached and circled the team’s boat, while responding in a conversational style to the whale ‘greeting signal.’ During the 20-minute exchange, Twain responded to each playback call and matched the interval variations between each signal.

I've long thought it would make sense to attempt communication with the intelligent species on our own planet! 

19 November 2023

Evolution and Free Will

Pulled from the blog list, the recent Brain Science podcast with Kevin Mitchell is worthwhile.

As with his new book, it's titled "Free Agents: How Evolution Gave Us Free Will" and was posted Oct 27, 2023.

02 May 2023

AI reads the brain?

Well, long time no posting!

The post "A.I. trained to read minds and translate private thought to text via brain scans" from BoingBoing caught my eye. Here with extensive training on a specific person's brain activity while listening to spoken text, is able to correlate later brain activity (while watching silent films or thinking of speaking) and do pretty well at reconstructing at least some of what the person "had in mind". Note though that patterns for one person do not carry over to other people.

This language-decoding method had limitations, Dr. Huth and his colleagues noted. For one, fMRI scanners are bulky and expensive. Moreover, training the model is a long, tedious process, and to be effective it must be done on individuals. When the researchers tried to use a decoder trained on one person to read the brain activity of another, it failed, suggesting that every brain has unique ways of representing meaning.

14 January 2018

Worms re-grow brains with old memories?

How much do we really know about memory storage?  This story from National Geographic may make you think again: "Decapitated Worms Re-Grow Heads, Keep Old Memories" by Carrie Arnold (dated July 16, 2013).
After the team verified that the worms had memorized where to find food, they chopped off the worms’ heads and let them regrow, which took two weeks. 
Then the team showed the worms with the regrown heads where to find food, essentially a refresher course of their light training before decapitation. 
Subsequent experiments showed that the worms remembered where the light spot was, that it was safe, and that food could be found there. The worms’ memories were just as accurate as those worms who had never lost their heads.

16 May 2016

What do we really know about Matter?

Two recent articles hit a similar theme, pushing the notion that our experience (consciousness) is in some sense on firmer ground that our understanding of physical matter.

The first I came across today via Twitter: "Consciousness isn't a Mystery: It's Matter" by philosopher Galen Strawson in the New York Times (May 16, 2016).  Here's the gist:
... we know exactly what consciousness is — where by “consciousness” I mean what most people mean in this debate: experience of any kind whatever. It’s the most familiar thing there is, whether it’s experience of emotion, pain, understanding what someone is saying, seeing, hearing, touching, tasting or feeling. It is in fact the only thing in the universe whose ultimate intrinsic nature we can claim to know. It is utterly unmysterious.

The nature of physical stuff, by contrast, is deeply mysterious, and physics grows stranger by the hour. (Richard Feynman’s remark about quantum theory — “I think I can safely say that nobody understands quantum mechanics” — seems as true as ever.) Or rather, more carefully: The nature of physical stuff is mysterious except insofar as consciousness is itself a form of physical stuff
I think this is on the right track...  emphasizing the primacy of experience, but not claiming that experience is necessarily exposing the actual nature of 'physical stuff'.  It's easy to assume we have a good handle on Matter, when in fact we've only discovered some rules about it, along with the working assumption that whatever it is, if you get a complex enough organization you get what we think of as conscious experience.

Back in April, Amanda Gefter wrote on and interviewed cognitive scientist (and author of Visual Intelligence) Donald Hoffman in an article entitled "The Case Against Reality" in The Atlantic.  Hoffman argues that our evolutionary path driven by fitness means that we have no reliable means of accessing what's really out there.
The idea that what we’re doing is measuring publicly accessible objects, the idea that objectivity results from the fact that you and I can measure the same object in the exact same situation and get the same results — it’s very clear from quantum mechanics that that idea has to go. Physics tells us that there are no public physical objects. So what’s going on? Here’s how I think about it. I can talk to you about my headache and believe that I am communicating effectively with you, because you’ve had your own headaches. The same thing is true as apples and the moon and the sun and the universe. Just like you have your own headache, you have your own moon. But I assume it’s relevantly similar to mine. That’s an assumption that could be false, but that’s the source of my communication, and that’s the best we can do in terms of public physical objects and objective science.
Presumably since all humans are on the same evolutionary path, we do indeed have similar assumptions and experiences. But it may be much harder to communicate with beings coming from different evolutionary pressures.

21 October 2015

More news on the worm C. elegans - a few more neurons?

 Been awhile since I've been active here, but this exciting worm news certainly rates a post.  To get caught up, check this previous post: Modeling the Worm!

In this story from Nature, "Surprise 'mystery' neurons found in male worms" the title gives it away.
The neurons help the worms learn when to prioritize mating over eating, revealing how a seemingly simple brain can be capable of a complex learned behaviour — and one that differs between the sexes.

 elegans worms are the model animal of choice for many neuroscientists, because their neural circuits are so simple that they can be mapped in full. They have two sexes: hermaphrodite and male. Hermaphrodites, the best studied, have just 302 neurons, but males have more — the MCMs raise their total to 385 neurons.
So it looks like there's more work to be done to get a good handle on the worm brain.