AI Reading comprehension scores

In September 2019 there was a great many blog posts about AI surpassing human reading comprehension scores.
Humans comprehension of a piece of text is 89.452%

As of June 4th 2021, AI is now hitting scores of 93.214, by IE-Net (ensemble) - RICOH_SRCB_DML

Plotting the scores on a graph:


Extrapolating the scores, we should see an AI hit 100% sometime in 2021.
Spreadsheet I created for the graph above: AI-Comprehension-score.xlsx

Original data source: https://rajpurkar.github.io/SQuAD-explorer/
Also see https://paperswithcode.com/sota/question-answering-on-squad20

Popular posts from this blog

Execution of user code in the .NET Framework is disabled

I was testing our new SQL Server 2008 enterprise cluster today and managed to get the following SqlException running a C# application: "Execution of user code in the .NET Framework is disabled. Enable "clr enabled" configuration option. The statement has been terminated." The solution is to run the following: EXEC sp_configure 'show advanced options' , '1' ; go RECONFIGURE ; go EXEC sp_configure 'clr enabled' , '1' go RECONFIGURE ; -- Turn advanced options back off -- EXEC sp_configure 'show advanced options' , '0'; GO
Read more

Service Broker sys.transmission_queue clean up

If you’re developing a Service Broker application and you misspell your services, forget to create a master key, or do any of the many things that result in your messages not being delivered, then your transmission queue will fill up. Due to the nature of service broker, errors don't occur at the client but rather your messages go in to Service Brokers sys.transmission_queue and are then subsequently processed. If they fail, they stay in the queue. Some errors can be resolved, i.e. missing services. However, others can't, or are just plain difficult. I.e. You've misspelt something, or not used the correct security. To clear your transmission queue you need to end the conversations, this is a handy little script for doing this. Be aware this ends ALL conversations in the queue and so should NEVER be used on a production system. To see how many messages are in your transmission queue, use this SELECT   COUNT ( 1 )   FROM   sys . transmission_queue On my system i7, D...
Read more

What do I use to write software?

I started off with Dbase III, then Borland Turbo C for about 5 years, then when Microsoft Visual Studio v1.0 came out, I moved to that. I've been using Visual Studio ever since. Coding in C++ and have since moved to C# when that came into existance. Do I use component? Yes. Which ones then? Telerik http://www.telerik.com/ Dundas http://www.dundas.com/ Rebex http://www.rebex.net/ I've also used DevExpress reports, but they peg the CPU to 100% on the webservers when exporting an excel spreadsheet with a lot of data in them, and more often than not, cause a timeout on the ASPX page. So I don't like to use them anymore. My essential toolkit consists of: SVN - Subversion source control http://subversion.tigris.org/ Winrar http://www.rarlab.com/ Colour Schemer (web design) http://www.colorschemer.com/online.html SQL Prompt http://www.red-gate.com/ ANTS Profiler http://www.red-gate.com/ NUnit http://www.nunit.org/ Microsoft Visual Studio http://msdn2.microsoft.com/en-gb/vstudio/de...
Read more