For predictions and the simulator, just scroll down. Note I am changing the whole script and building a new approach, which will ensue two methods of calculations. The current script is a complete re-write from the previous (as of January 4th 2017) one, where it now starts plotting from the estimated singularity (it tests years from 2026 to 2060) and follows backwards until the stop condition to match the historical trend. The new script (yet unreleased) will use a conventional fixed-time approach.
For the easy answer, here it is: 2026~2028.
Using the history database and a script that detects the exponential line that best suits the speeding technological advances seen so far, and supposing this "accelerating returns" keep going in the near future at least until the development of the most fundamental AGI module, one can predict with a good margin of accuracy a date when the exponential growth tends to infinite, which basically means that if the script kept simulating the years, the one year next to the detected would have the same amount of technological breakthrough than all history before.
Since there are thousands of simulations to be done for every single assumption made (when technological history starts? how fast is the acceleration, in which year to lock the possible singularity to detect the logarithmic curve end?), there are some pre-made predictions below. Then you can choose your own assumptions and check how the simulation goes (might be slow unless on a high-end computer)
I am yet to write a proper essay on how this script works, the internal choices and the input parameters, but that would be a long text and this is not the meaning of it. So, here is a small summary on the possible choices you have when setting the prediction tool:
"Simulate with data up this year": You can simulate as if in the past (without current knowledge). The simulator will stop feeding on data from the database starting on the selected year. Default: this year.
"Simulate as far back": The script conjectures the singularity date and then plot the evolution backwards until this date. In theory, it could go as far as -100.000 which is what data we have on the database, but that would be irrelevant because before written language, we have no solid records of evolution.
"Speed at which scientific progress accelerates between": For a long time, scientific progress was thought to be scalar, but it has been since been proved exponential, as new discoveries help and speed the development/discoveries of new knowledge. The script can find which will fit better, and will also show you the best match for every year other than the best match.
"Slow down accelerating returns/Moore's law": While it has been proved that for at least the last 70 years we have accelerating development, a lot of people see in Moore's Law breaking point an argument to say this is going to stop. Choose this to reduce future accelerating returns
"Total break-down of accelerating returns/Moore's" Law: As the name implies, choose this to stop all acceleration and revert back to scalar evolution.
"AGI theorem of self-development": It has been widely argued that and General Intelligence would offer self-improvement, since the definition of AGI is that it can learn and improve by itself. Choose this option if you believe that, once even the most basic AGI is developed, it will accelerate it's own development if compared to humans alone (tandem research). This will affect only the development after predicted AGI is created, which is part of the script.
"ASI theorem of self-development": Just like AGI will speed up development of itself, it will also speed up development of ASI (SuperIntelligence) so it is reached faster than we predict. The time between full AGI and ASI is hardcoded as 5 years, and then reduced per evolution factor.
"Lock only technology progress": Consider only technological progress for the singularity is required. Remarkable but not I.T. related developments on the history database will not be considered.
"Compensate numbers for past": Since even after written language started we lack a lot of records on developments and discoveries, the code will compensate the Development Index of all years before the year 0 gradually until a maximum compensation at -3000, at which point it stays the same compensation. This compensation is a percentage, not an absolute number, over existing discoveries.
Once you are ready, just click "Start Prediction"
1. What the simulation does when you switch on/off the accelerating returns, AGI or ASI? - Nothing. The data we have up to now is how technology progress without outside impact, and so far accelerating returns is ongoing. The change is calculated at the end results, as follow: Given the date result, 5 years is removed for the ASI progress, then 60/f years is removed for the AGI process, where 60 is an approximate interval of years between the advent and apex of computers (1945~2005), divided by the accelerating returns factor (so, if it too 60 years, and we are accelerating at a factor of 2, it would take only 30 yeas now). The resulting year is the calculated start of Singularity, where base AGI is expected. If Moore Law's totally fails as predicted (in 2017), the difference between this year and 2017 is added (because the expected boost by acceleration is no longer present), or this difference divided by f if you choose "reduction" instead of total break down. Now, AGI helps by speeding the singularity by 60/f2 (note the power of 2, because one f was already applied naturally), and again ASI boosts by 5/f.
Please check the source code of this page for all the scripts and database. It is Open Source - Copy Left.
Fill this with your estimates. The default values are my predictions.
Simulate with data up this year:
Simulate as far back as:
(-100000 ~ 1900)
Speed at which scientific progress accelerates between*:
(1.01 to 4)
IMPORTANT: mark only if you believe that these events will make progress leave normal curve. If you believe they are part of progress (highly likelly), leave all unmarked.
Estimated starting 2017, slow down progress
Stop speed up completely, becomes scalar
AGI helps its own development
ASI helps its own development
Ignores developments not relatable to IT/AI/AGI/ASI
Simulations stop at -3000 due to too few records to properly calculate the Development Index. However, there were also not much written up to circa 1000. Mark this to try and compansante that
If you are skeptical for how early singularity is, keep this checked for skeptical calculations