9 Matching Annotations
  1. Aug 2020
    1. Imagine you prompted the model with “What is one plus one?” I actually don’t know how it would do on this problem. I’m guessing it would answer “two”, just because the question probably appeared a bunch of times in its training data. Now imagine you prompted it with “What is four thousand and eight plus two thousand and six?” or some other long problem that probably didn’t occur exactly in its training data. I predict it would fail, because this model can’t count past five without making mistakes. But I imagine a very similar program, given a thousand times more training data and computational resources, would succeed. It would notice a pattern in sentences including the word “plus” or otherwise describing sums of numbers, it would figure out that pattern, and it would end up able to do simple math. I don’t think this is too much of a stretch given that GPT-2 learned to count to five and acronymize words and so on.

      This is also borne out in my own tests. Easy calculations, the likes of which the model must have seen or easily learnt, it does well on. More exotic ones not so much.

      What is interesting is that what predicts whether or not GPT3 is able to do the calculation is not the difficulty of the calculation, but the likelihood it occurred in its training.

  2. Jul 2020
    1. LoD = LoB + 1.645(SD low concentration sample)

      LoD is the lowest analyte concentration likely to be reliably distinguished from the LoB and at which detection is feasible. LoD is determined by utilising both the measured LoB and test replicates of a sample known to contain a low concentration of analyte.

      LoB is the highest apparent analyte concentration expected to be found when replicates of a blank sample containing no analyte are tested.

      LoB = meanblank + 1.645(SDblank)

  3. May 2020
  4. Apr 2020
  5. Feb 2020
  6. Apr 2018
  7. Oct 2013
    1. When, however, something is to be done, and we are speaking to those who ought, but are not willing, to do it, then great matters must be spoken of with power, and in a manner calculated to sway the mind.

      The word calculated carries overtones of manipulation without emotion or concern