Casually explain what HTM is and what AI research is doing

Hi all in the forums.
Some of my friends are asking me what am I doing at the weekends. Well… I have been messing with HTM algorithms and neural networks. Yet, how can I explain this in a not nerdy way? I need suggestions.


Perhaps an impossible task? :sweat_smile:


Give up on not sounding nerdy - embrace it with stuff like:
I am researching brain circuitry.
I set up computer simulations of key parts of the cortex; so far the results are encouraging.

Not good enough? Add this bit and watch the eyebrows go up:
I am inventing a program called (your name here - sky-net is too obvious). I have been feeding it a training set of advanced hacking techniques. Like any deep learning system it is not clear exactly what the internal model does but it is clearly working; just this week it looks like it got into some sort of military networks. I never thought it would work this well!


Is the goal not to sound like a nerd, or to explain it convincingly to someone who is not a nerd (but hopefully doesn’t mind you being a bit of a nerd)?

Note: quite a few of my friends considering being a nerd as a badge of honor. :-).


Great question! (the importance of a good creative decision may go without saying)

To your dilemma; if results may speak for themselves, then perhaps the audience may engage on their own terms?

I think you missed the main thrust of the question.
So - your co-worker asks - what are you working on at home and your answer is: my work speaks for itself?

Casually explaining the rationale behind biologically constrained ML research is pretty easy.

Casually explaining the components of HTM theory is pretty hard.

I normally do the former, then point them to materials that do the latter (though even HTM school isn’t really casual)

1 Like

Explaining to my co-workers is easy. I usually say “A stripped down version of Spiking Neural Networks that uses less resource but preserves the behavior” and they know that NNs are.

I’m trying to not sound nerdy.
It’s a personal thing. Unfortunately though I enjoy a nerd. Being nerdy is one of the best ways to push general friends away in my country. It sucks, but reality. And being nerdy is one of the best ways to not get into any relationship. :frowning:


Nicely done, Mark!

Your example may demonstrate a distinction between how to “explain” it (Marty1885), versus what to say or what to call it when someone asks, and although such a thing may well entail some level of verbal or written response as you may be suggesting, one might interpret the request as a matter of how to best present the subject matter to a general audience.

Gary Gaulin for example, refers to the character featured in one of his spacial map releases as a “critter” and when Gordon Sumner was asked about what it was he was referring to in the song If I Ever Lose My Faith in You, responded with any number of examples pointing to none in particular.

Our spacial map example might be described as a “directionally oriented, topographically organized neural map of the spatial environment contained by the dorsocaudal medial entorhinal cortex” under Moser (2005), but in general terms, “figure” may be a better choice. (general versus specific?)

Perhaps the author of the play stated it best in the Act II Scene II?

“” A rose by any other name would smell as sweet " is a popular reference to William Shakespeare’s play Romeo and Juliet , in which Juliet seems to argue that it does not matter that Romeo is from her family’s rival house of Montague, that is, that he is named “Montague”. The reference is often used to imply that the names of things do not affect what they really are. This formulation is, however, a paraphrase of Shakespeare’s actual language. Juliet compares Romeo to a rose saying that if he were not named Romeo he would still be handsome and be Juliet’s love. This states that if he were not Romeo, then he would not be a Montague and she would be able to get married with no problem at all."

“A literary technique, originally used in Greek tragedy, by which the full significance of a character’s words or actions are clear to the audience or reader although unknown to the character.”

In my opinion always say “HTM Theory” not “HTM”.

And never say “AI research” that can now sound like kid stuff. Numenta more specifically uses a “computational neuroscience” based approach to understand how the human mind works.

You’ll maybe still be categorized as a “science nerd”, but at least your area of study will sound more impressive. With luck you too might be honored by making it all the way to “mad scientist” level.


I feel your pain. :-7

The best way to explain something is to project yourself at the level of knowledge of your audience. (That’s even more important than your own understanding of the subject. Teachers often forget this). So you have to use the vocabulary of the person(s) you’re talking to.

A good trick is to use fitting metaphors. (I use the concept of a balance scale to explain action potentials, writing on squared paper to explain anchoring to grid cells, …)

Another thing I do as much as possible is only answer questions. As long as they ask questions, you know they are interested (or at least polite). As soon as the questions stop, you’ve reached that dangerous line of becoming boring. Also, when someone interrupts you with an unrelated topic, do not pick up the previous conversation unless they ask you to continue. (But in my experience that rarely happens. Another telltale).

If you absolutely must continue the conversation, then ask questions yourself. Let them think about consequences or thought experiments. That keeps them involved.

A last idea is to try to learn something from that person yourself. Anything. Ask their opinion on problems you’re stuck with. (I have to admit though, that that is something I don’t do enough myself).


Embrace the nerd! :sweat_smile:

To my non-nerd friends I say I’m working to make the world a better place by changing the way alarms work in intensive care (i.e. the why).

To my medical friends (who may or may not be nerdy) I say I’m using principles found in how we process biological sensors to detect physiological changes in critically ill patients (i.e.the what).

To my nerdy friends I say I’m developing an implementation of a neural network in R to detect anomalies in real-time physiological data streams (i.e.the how).

Always be prepared to modify your delivery to suit your audience :blush: