How do you get new ideas?


How much time should we spend exploring alternate ideas?

In the roaring flood of ideas and papers that are available, it would be easy to dissipate all of your energies just trying to plow through it all. It takes time to understand and weigh the ideas. Frequently the ideas are poorly presented and/or use unfamiliar language so it veers off into the realm of hard work to even understand what the author is trying to say.

That said - these are also the source of new ideas that may eventually come to be the best way of understanding the topic. I struggled with SDRs for a while but stayed with it and now see it as a powerful tool.

I have also spent perhaps more time than I should looking through stuff that I eventually concluded was a waste of my time. This makes me wonder if it was a bad idea or just the cost of searching for new ideas.

I am sure that many paradigms start out as uncomfortable ideas that take time to absorb. Some examples: relativity, evolution, plate tectonics. Many in the related fields went to their deathbeds convinced that these were just plain wrong.

Do you let others vet the ideas for you? (references in papers, pointers from people you trust?)
Do you scan through and stop as soon as you hit something that seems "off?"
Do you just ignore anything that is not in the middle of the accepted paradigm?


I cycle between “research” and “production” phases when learning new ideas. I have a hard time internalizing new concepts without building something (hence all the visual examples in HTM School).

When approaching a new idea from the outside, I usually have a long research cycle that involves reading and reading and also reading. Once I feel like I have informed questions about the topic, I’ll go ask them to people I know will help out. As soon as I think I have the main ideas of the concepts, I’ll start building something and immediately hit a brick wall because I misunderstood something. :slight_smile:

Then the cycle starts again.


Personally, I think the only way I have any chance of contributing to HTM is by looking for alternate ideas, or at least researching an alternate topic. Anything Numenta is currently working on is something I mostly avoid researching because I want to avoid redundant work and it will take me way longer to make progress.

Even for basic things like how different types of cells connect, there aren’t many solid facts. To deal with this, I try to integrate results from different papers. If a bunch of papers with different methods support the same result without obvious flaws, that result is much more certain and patterns can start emerging which hint at functionality.

I’m not sure how well that approach applies to ideas, though. I catch a lot of misunderstandings when I go back and boil down my notes on an article a different day than I took those notes. It might help with the unfamiliar language if you stick to certain regions, species, and/or modalities.

More broadly, it helps me if I approach a topic like a detective case. Finding contradictions or opposing ideas helps me. Researching things like cognitive biases and researching techniques for scientific research (e.g. keeping multiple hypotheses at once and not favoring papers I read first) also helped me.

I spent several months researching a topic (ramping activity/presaccadic remapping) which I ended up dropping. I stubbornly kept researching it for a while even though my progress grinded to a halt. Because I don’t want to repeat that, I avoid things without a firm basis now. I still take notes on that sort of thing when it interests me, but I don’t explore those topics beyond the first article. I don’t know if this is the right approach, of course. Maybe I’ll return to that topic once I’ve got a more solid foundation.

Still, I throw out my notes for some articles, and a dozen articles usually boil down to a page or two of notes. It takes a while.

Finding these paradigm-shifting ideas among all the poorly written articles is hard. It might be faster to arrive at the right paradigm by finding problems with the current paradigm.

When an article cites a fact, I check the cited source because a wrong fact could throw everything off, but I generally don’t care enough to keep that cited fact unless I want to read the cited article anyway.


4 posts were split to a new topic: Think about nature