Wednesday, May 25, 2016

Imagine Discovering That Your Teaching Assistant Really Is a Robot

You may remember that back in the mid 1990s IBM created a computer system, called Deep Blue, that challenged Grand Master Gary Kasparov at chess. Although initially unsuccessful eventually Deep Blue beat Kasparov. This was a great achievement but there was no other use for the program; Deep Blue could play excellent chess but that was all. It couldn't be repurposed to any other task. Essentially Deep Blue was a dead end. So in 2011 when IBM set itself another challenge to develop a computer system, called Watson, to answer questions on the quiz show Jeopardy they already had future uses in mind. Watson was intended to answer questions in any domain where there might be a large body of written knowledge such as medicine or the law. The Wall St Journal recently reported on an interesting use of Watson as an online teaching assistant, called Jill Watson, at Georgia Tech. "Online learning has opened the door for instructors to reach a gigantic, global audience. Trouble is, many students can ask a lot of questions, burdening human TAs, said Ashok Goel, a professor of computer science at Georgia Tech. It was he who recruited Jill Watson for his Knowledge-Based Artificial Intelligence class." Perhaps the most interesting thing about this is that many students assumed Jill Watson was a person. Has the Turing Test been passed?
Thanks to my colleague, Mark Wilson, for bringing this story to my attention.

Monday, May 23, 2016

Free public lecture on computer science

The final lecture in the series of free public Gibbons lectures takes place this Thursday 26th May at 6:30pm. This week's lecture titled: Using IT to improve health delivery for management of chronic illness is given by my colleague Professor Jim Warren, Professor of Health Informatics, Department of Computer Science, The University of Auckland. It has been suggested that more could be done to improve human health by better delivering what we already know, rather than by developing further medical knowledge. There are enormous gaps in health delivery, including inadequate detection of individuals at high risk and mismanagement of risk factors once they are identified. IT can help through the use of decision support systems that compute estimates of risk and recommend appropriate management. IT can also serve to make screening for risk factors more systematic, and consumers can receive direct 'eTherapy' for problems such as anxiety or to aid in quitting smoking.
More details about the talk, time and venue are here. If you were unable to attend the previous week's lectures they can be streamed from here

Thursday, May 19, 2016

Peter Norvig, Research Director at Google, answers some questions

One of the great things about teaching AI today, which I do, is that our standard textbook, AI: A Modern Approach, is co-authored by Peter Norvig who is Research Director at Google. You really can't show how important AI is today than that. My colleague, Mark Wilson, brought this interview with Peter Norvig on Quora to my attention; recommended.

Monday, May 16, 2016

Free public lecture on computer science

The third in the series of free public Gibbons lectures takes place this Thursday 19th May at 6:30pm. This week's lecture titled: A case study of IT in Medical Imaging: The evolution of Computed Tomography is given by Professor Anthony Butler, Head of the Department of Radiology and Director of Centre for Bioengineering & Nanomedicine, University of Otago at Christchurch. When x-rays were first discovered, medical images were viewed with devices such as fluorescent screens and eventually from x-ray film images. In the 1970’s, with the introduction of computers, 3D x-ray images could be produced using machines known as Computed Tomography scanners, or simply CT scanners. More recently, by adapting microchip technology for use in x-ray detectors, it has become possible to measure the x-ray colour (or spectrum) in CT. This facilitates the measurement of tissue constituents that were previously difficult to discern. Medical researchers are applying this new knowledge to problems in vascular disease, cancer, and joint disease. This talk outlines the development of medical x-ray technology, highlighting advances made by NZ researchers over the last 45 years.  More details about the talk, time and venue are here. If you were unable to attend the previous week's lectures they can be streamed from here

Thursday, May 12, 2016

Building AI Is Hard—So Facebook Is Building AI That Builds AI

Wired reports that "deep neural networks are remaking the Internet. Able to learn very human tasks by analyzing vast amounts of digital data, these artificially intelligent systems are injecting online services with a power that just wasn't viable in years past." Read the full article here.

Monday, May 9, 2016

Free public lecture on computer science

The second in the series of free public Gibbons lectures takes place this Thursday 12th May at 6:30pm. This week's lecture, titled: Finding your place in the genome: assembly, annotation, association, is given by Prof. Thomas Lumley of the Department of Statisticsat  The University of Auckland. To an extent that would have been astonishing a century ago, DNA copying and transcription turns out to be a digital process. The copying mechanism allows tiny amounts of DNA to be amplified; the base-pairing of the double helix lets us read out a sequence, usually by adding fluorescent tags to the DNA bases. With three billion letters of genome, much of which is poorly understood, computers need to do nearly all the work. The talk will discuss three areas of IT involvement - DNA sequencing, annotation (looking up what is known or can be guessed about a stretch of genome) and association studies (relating differences in the DNA sequence to differences in biology and health). More details about the talk, time and venue are here. If you were unable to attend last week's lecture it can be streamed from here

Thursday, May 5, 2016

You Pay to Read Research You Fund. That’s Ludicrous

Wired has recently run a piece titled You Pay to Read Research You Fund. That's Ludicrous. I've blogged about this before and I know that it's a subject that many of my colleagues are concerned about. Really you'd think that a bunch of smart people (aren't academics supposed to be smart) would come up with a better solution. Back in the days when scientific journals had to be printed and posted out to institutions and subscribers it made sense to let publishers do this, and they deservered to be paid for their service. Now that all journals are available online paying publishers to do this is crazy.