As one of those who was fortunate to be at the conference to hear Steve Alessi's Dean Lecture, I must say it was most thought-provoking. Here are a couple of relatively minor comments about the opening section, which was largely a piece of enjoyable crystal-ball gazing, plus some thoughts about the nature of courseware. If you like your discussion academic and esoteric, read no further.
Coincidentally only a few days after the conference, I was talking to a teacher who specializes in supporting the learning of deaf and hearing-impaired children and young adults. These children already suffer from disadvantages when it comes to learning to read and speak. While it may be possible to create an interface where members of that group can receive an output from a new electronic reading device (NERD?), perhaps in signs, being able to use a speech input device may always be beyond them. In Steve's future becomes reality, some disadvantaged groups may find themselves even more excluded.
My second point concerns the notion of a universal translator. Translating language is difficult enough, but that's not the whole story. Having had experience of translating courseware between languages, including American into English, it is really meaning that needs to be translated and that is a far trickier proposition. (See my paper to the ADCIS conference 1989, International Courseware--The Problems and Pitfalls). In any case, Steve's assertion that "most people do not learn a foreign language as it is" is very sweeping. Make that "many people who have English as their mother tongue do not learn a foreign language." The story is certainly different in mainland Europe, where it seems that most people have some proficiency in more than one language.
Now to cooperative/collaborative learning. Even where courseware is designed for individual use, schools often put two or more students in front of each computer due to the limited numbers of machines. I can see great merit in more research into how to exploit that for improvements in learning. This is fine for schools. For my business customers, who are using CBT to offer employees open learning opportunities at the work-place, this is less likely to be of use. The main topics are work-related procedures and operations, with on rare occasions where students are in neat groups. The prime goal is the maximum learning in the minimum time. It's difficult to see a place for anything but individualized training. This is a conservative view, but businesses are like that.
Finally, Plenty-of-Time versus Just-in-Time learning. Just-in-Time training and its illegitimate sibling EPSS are fine when (1) learners don't need much practice to be proficient and (2) only small numbers of learning points have to be acquired to complete the task. Airline pilots and surgeons are not high on my priority list for these types of learning. I have a pragmatic approach here. With most types of teaching we do not expect students to learn everything that is in the lessons (or all post-tests would require 100% to pass). In particular, most job-related training offers an opportunity to learn, but we must expect trainees to return to the job incompletely trained. Usually a colleague or supervisor fills in the gaps. That's what really happens, and it has many advantages. Just-in-Time learning bypasses the settling-in and learning-completion phases which are arguably essential. Plenty-of-Time Incomplete Learning is a more accurate description, with people-based performance support the norm.
I offer no apologies for bringing this down to earth and into the workplace.