About that computer with 150 IQ


I read the short Rand paper on Delphi Method by Olaf Helmer from 1967. Summary of Olaf’s version:

  1. Create a model for a real-world problem, decoupled from the actual problem, but with the same moving parts
  2. Ask experts to interact on the model and then individually opine on the outcome
  3. Go through some iterations of voting, maybe you’ll get one consensus opinion, or maybe two
  4. Weight the votes based on prior demonstrated competence of the individual experts in similar questions

Helmer gives an example of forecasting when a computer would achieve an IQ score of 150. Based on Delphi method, the median expectation in 1967 was 1990, with an upper bound of 2000.  I did some Googling on this topic, and the basic answer is: hasn’t happened yet. There was a Swedish professor who made the claim in 2012:

However not much as been heard in the news from him since then.  He does, however, have some very interesting papers on his website.

It turns out that Chinese at Microsoft are the closest to actually getting that done. Interestingly, in both the Swedish Professor case and Microsoft, they use Amazon Mechanical Turk performance as their benchmark in terms of “beating humans” at the test:

In thinking about measuring artificial intelligence, we must be careful in this regard not to fall into the Watson trap identified by Roger Schank:

although Roger has his critics:

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s