Recommendations Are Far From Good

Doctors used to recommend leeches to cure a whole variety of illnesses. That seems laughable today. But I think our recommendations today will be laughable to people in the future.

Recommendations work terrible for everyone but decently on average.

We are a long, long way from making good individual recommendations. You won't get good individual recommendations until your individual genome is taken into account. And even then it will take a while. We may never get to the point where we can make good individual recommendations.

So many cures and medicines work for a certain percentage of people, but for some people they can have detrimental or even fatal effects. People rave about certain foods, exercises, and so forth, without considering how differences in genetics can have a huge role.

People are quite similar, but they are also quite different and react to different things in different ways. I think we are a long way away from seeing breakthroughs in recommendations.

Recommendations are great business, but I think we're 2 or 3 orders of magnitude away from where they could be, and it could take decades(or never) to reach those levels.



Posted 08/25/2010

Was this essay useful to you? Yes | No
Powered by
brecksblog Posts: What can a Programmer learn from Rock Climbing? Look for a Line Backpack the World with Zero Planning The Economy Explained Ruby You Can't Predict the Future Critical Thinking Kids are Neat How is Intelligence Distributed? Recommendations Are Far From Good Nature Verse Nurture Circle of Competence What Percentage of the Brain Does What The Recency Effect The Ovarian Lottery & Other Side Projects Happiness is in Mediocristan What I Want Orbits The Do You Know Game and Why We Need Celebrities Design Matters, a lot Competition and Specialization Simple, but not easy Flip Flopping The Churn Rate of Data Culture and Complexity The Invention of Free Will Why is it best to do one thing really, really well? The Hidden Benefits of Automation Metrics for Programmers HackerNews Data: Visits as a Function of Karma Don't talk about what you will do, talk about what you have done Why it's worth it to buy the book The Least You Can Do Four Tips to Improve Communication Network Effects If you can explain something logically, you can explain it simply With Computers: Don't Repeat Yourself. With People: DO Repeat Yourself When forced to wait, wait! How to Buy Low, Sell High Flee the Bubble Checklist for New Products Diversification in Startups Thoughts on Setting Goals Problems Worth Solving Make Something 40% of Your Customers Must Have SEO Made Easy: LUMPS Don't Flip the Bozo Bit (un)features Get Stuff Done By Setting Arbitrary Constraints Why You Shouldn't Save Blogging for Old Age 6 Specific Ways to Find Programming Mentors Orders of Magnitude The Many Worlds Law Twelve Tips to Master Programming Faster What's Linear Algebra? Fiction or Nonfiction? Experience is What You Get I'm Back Check out my roommates' companies:
Jellyfish Art makes great Jellyfish tanks
30 Words makes great language guides