Hacker News Reader

Comment by hprotagonist | Hacker News Reader

It is critical to note that this study does not address the transfer of learning or performance on one task and show that it improves performance (or in this case, energy efficienty) on any other task.

Put another way: lumosity makes you good and efficient at lumosity, but not anything else.

Transfer learning is a bugbear for neuroscientists as well as ML people, sadly.

View on HN

Totally anecdotal story, but from my own experience, I agree with you completely.

I was using one of the leading 'brain training' apps, and they started requiring a subscription so I switched to another app. Despite having a strong performance in the first app, I had quite a weak performance initially in the second app.

In other words, skills aren't even transferable between brain-training apps, let alone into real life applications.

Parent is highly informed re: transfer. Another 'bugaboo' of cognitive training is the opportunity cost of training. For example, instead of cog training, use the time to get cardiovascular exercise, which has proven and robust benefits for cognition.

Isn't that exactly what the experiment did? They divided subjects to 3 groups: 1 with cognitive training, 1 without anything in particular (wait list), and 1 with physical exercise.

I hadn't realized because I only stopped in to reply to the grandparent's post, but that is fantastic! A really excellent active baseline.

Baseline for activity is too low - 30 minutes five days a week.

To prevent brain tissue loss one should walk 35km per week, it is one hour daily. And that amount is to stop decline of mental capabilities, one need to increase that amount of exercise or add fasting or weight training to start to regain "younger" brain.

So I am not impressed.

I can't access the study text, but if anyone can could you check if the task they used to measure results was really the same task that they trained the participants on? It seems like thats an obvious flaw that they would have thought of.

http://www.neurobiologyofaging.org/article/S0197-4580(17)303...

if that helps..

Is there any kind of learning that improves performance on other tasks? (What about language? What about if the 'other tasks' include learning?)

There is some preliminary evidence that a training task called "dual n-back" may improve working memory performance. Good working memory is beneficial for a wide variety of cognitive tasks.

http://brainworkshop.sourceforge.net/

Learning to play an instrument and play music and learning a foreign language are both supposed to be quite good for you, mentally speaking.

There is some evidence that meditation can improve cognitive performance.

I was actually just looking for scholarly research on this subject, link?

If you dig around, you'll find a lot of studies that are focused on a particular outcome (e.g. effects of mindfulness on the Stroop test, effects of mindfulness on athletic performance.) The quality of these studies is, of course, all over the place. Most of the research seems to come from journals dedicated to mindfulness in some capacity. For example, would you trust a study done by PhDs at reputable universities if they published in "The Journal of Alternative and Complementary Medicine"[0]?

The best research I ever found was a metastudy in Clinical Psychology Review: "Does mindfulness training improve cognitive abilities? A systematic review of neuropsychological findings"[1] concludes that there's enough evidence that further investigation is warranted.

There's also a lot of downstream research (citing articles) from both publications that are worth investigating.

Transferability though, is still an open question. Even in this meta study, the author's admit that certain types of mindfulness training may be beneficial to certain outcomes.

So perhaps you'll be more mindful and less distracted, but you won't be any "smarter."

[0]: http://online.liebertpub.com/doi/abs/10.1089/acm.2007.7022 [1]: https://www.sciencedirect.com/science/article/pii/S027273581...

So does the same type of thing apply to tech-interviews vs actual on job problem solving capabilities?

It's the problem of studying for the test vs studying to learn. Some knowledge will transfer, but the further the test is removed from reality, the lesser the transferable knowledge will be. When I look back at the interviews I had and what I actually needed to know at work, the difference is massive.

One important way that all tests are removed from reality is that the direct motivation for solving the test problem is to score well on the test rather than to understand the problem better. In real life we are sometimes motivated by factors such as curiosity or survival which marshalls our internal resources more helpfully.

(And maybe evolution gave us curiosity to save on battery power with the new knowledge it produced!)

Reply on news.ycombinator.com