What I like about AI: conceptual fusion. The train biologist

To conceptually transplant a biologist into the role of a "Train Biologist," we must treat locomotives and rolling stock not as mechanical objects, but as a diverse phylum of living organisms. Here is how the core roles, subfields, and methodologies of a traditional biologist map to a Train Biologist. 

Evolutionary Biology & Taxonomy 

Cladistics & Phylogeny: 

Mapping the evolutionary tree of trains. Steam locomotives serve as basal ancestors, electric trains as highly evolved specialists, and high-speed rail (like the Shinkansen) as the apex predators of the phylum. 

Vestigial Structures: Studying defunct parts. A modern diesel-electric train carrying old, unutilized mechanical housing is analyzed the same way a whale's hip bones are studied. 

 Speciation: Examining how a single locomotive "species" diverges when isolated on different islands or continents with varying track gauges (allopatric speciation). 

 

Anatomy, Physiology & Pathology 

Internal Anatomy: Mapping the engine block as the heart, the fuel lines as the circulatory system, and the pantograph (for electric trains) as an organ for absorbing ambient nutrients. 

Metabolism: Analyzing how a steam train digests coal and water to produce kinetic energy, compared to the highly efficient, low-waste metabolism of a maglev train. 

Pathology: Redefining mechanical breakdowns. Engine overheating is treated as a fever; wheel-wear is studied as a degenerative bone disease; rust is a fungal infection. 

 Ecology & Behavioral Studies 

Habitat & Niches: Subway systems are studied as subterranean, nocturnal ecosystems. Shunting yards are viewed as dense breeding grounds or communal nesting sites. 

Symbiosis: Analyzing the relationship between the train and its human passengers or crew as a form of complex mutualism (humans maintain the train's health; the train provides migration). 

Migration Patterns: Tracking the seasonal or daily migration of commuter trains moving from suburban feeding grounds into dense urban cores every morning. 

Fieldwork & Conservation 

Field Biologist: Sitting hidden in a station or by the tracks with binoculars and a notebook, documenting the behavior, mating calls (horns), and territory marking of different models.

 Conservation: Working in railway museums to preserve "endangered" or "extinct" species (like old steam engines) and maintaining their genetic material (blueprints) for historical cloning (replications).

All the above is Google Gemini. What I enjoy is the abillity of large language models to think by analogy, something I've done for a long time as a private indulgence, but found little support for in formal academic literature.

The Entire world imagined through a Tesla

The entire world viewed from a Tesla. Elon Musk, stated today that his attempt at artificial general intelligence had an overwhelming advantage over other attempts, insofar as it was based on real world data, rather than, as he put it, the world as seen through a pdf lens. Source https://aisecret.us/musk-finds-the-only-road-to-agi/?ref=ai-secret-newsletter Its an interesting but underwhelming argument. He also claims using X, Neurolink, etc. as training material. The major human problem I see here is privacy. When I buy a Tesla, was I consenting to have my data used as part of a training set? When I opt for Neurolink surgery, assuming such a procedure becomes routine, did I consent to have my thoughts captured to improve performance of Grok 7.0 or whatever? AI ethics is complex, and very culturally nuanced. I can’t see where ethical training comes into the above. I stated that the above was the major human problem, because, even if you abstract human concerns from this project, the other major problem is how limited the rest of the data set is. Does X represent the sum of human discourse, or is it, as some people increasingly contend, an echo chamber for botnets to shout at each other https://en.wikipedia.org/wiki/Dead_Internet_theory And think of all the world off the beaten track, that Tesla’s don’t see. A lot of AI rolls back to good old statistics, and one of the first principles of statistical research is to get a representative sample. Not a sample of convenience. That step, is what is missing here.

 Nature

I've just read a report on how 20 minutes of Nature, three times a week are good for you, and how the benifits improve it the nature is wilder. that is Wilderness vs.calming city park.

I've thought for some time about a related concept, the idea of a field density of human decisions. I feel more comfortable, in a location where the density of human decisions around me is low.

Beaches, forests, mountains, and conversly I feel very stressed in cities, or crowds, where the decision density is high.

Just a thought.

 

The peacemaker Colt, and AI as guardians of Etiquette.

 My Cousin Nancy, sent me a poem, she had written on her phone. The phone refused to let her use the word "Damn", telling her that she was better than that.

A hundred years before, there was a theory that having an armed citizenry in the old, rosy and romantic west, ensured that it was a polite society.

Currently there are some areas of the US with very high levels of gun ownership, particularly among the criminal classes, but the evidence is still out, as to whether this has resulted in politer public life. I suspect it has not.

Returning to the phone enforcing standards of discourse. The work around it is to use a broader vocabulary, and perhaps regional terms, for which the phone is unable to pick up a double meaning. I can see language becoming increasingly local, and dependent on dialect, as phones take on a greater and greater role of policing word use.

More echo chambers and bird-whistles.

Once more the law of unintended consequences rears its ugly head.

Are we reading or Thinking

 I've just finished reading a short paper by Edmund Gettier. Its pretty much standard fare for any student of modern philosophy, and can be found by a trivial google search. His paper attacking the concept of Justified True Belief, was perhaps anticipated in part by Russel's example of Alice, and the Clock, but this is common in academia, as often a field is ripe for a new concept.

It did not cite Russell, though it did Cite A.J Ayer, whose work on the Problem of Knowledge had been published previously.

Both these papers are considered important in their field, and I think were considered to be so well known that they might not need citing, at that time. 

On a personal note, as in this is just my opinion, I think that as the universities have spread their wings to include more and more people, there has been a change in emphasis in academic literature away from thinking and original ideas, and towards citations and Authorities, in what is know as “Scholarship”.   This was the way of the schoolmen, and is similar to the Confucian approach to education.

These may not be the most fruitful of paths to follow for creativity and innovation.

Edmund Gettier Is justified True Belief Knowledge? Analysis. vol. 23 (1966).

A. J. Ayer. The Problem of Knowledge McMillan (London, 1956).

AI AI its off to work we go AI AiAi.

We sit on the cusp of an AI revolution, and the explosive development of ChatGPT and its ilk have finally

woken the media to the creeping presence of AI driven tools in everyday life. For a very long time AI was like controlled Nuclear Fusion, progress was just around the corner.

Nuclear Fusion has yet to become commercially viable, but AI is now sucking in money as if the word "bubble" never existed.

So perhaps a few thoughts here. 

AI is not one thing there are many different AI methods out there. They are behind recommender systems, speech recognition, autonomous driving, smart speakers and so forth.

Word is probably the most commonly used program which uses AI, without most people being aware of it.

AI challenges our ideas of putting together ethical frameworks for use. 

A framework which supports Transparency and explainability, and the right to be removed from the model, along the lines of European GDPR legislation, 

Has nothing to say about the thought of one million two inch long autonomous drones equipped with movement recognition and munitions, being shipped to an enemy port to enforce a curfew for a week.

Dystopias abound.

On the other hand the potential to use pattern recognition to synthesize new drugs or medicines, or the Idea of a mental pattern recognition pacemaker, which could intercept a seizure and close it down, is tempting.

So far, all the proposals for education I've seen have been very conventional, but perhaps a future deep learning algorithm could examine the practice of many teachers, and identify teachable moments, and work out whats the best way of maximising learning.

I've been teaching AI for the past two years, and first encountered it in 1982-3. During that time I have been privileged to encounter many humans who's mental landscapes are very different in the way they are wired or in their ability to interact with others. Many of these encounters have been in therapeutic settings.

Studying AI has given me insights into how many different ways of thinking there are, and into just how many ways humans differ from mental models.

This blog is called thoughts in transit, and I put it down as just that a location for ideas which are still travelling and which have not yet come together.

One final thought. In trying to understand the coming AI technological revolution, the closest model I found in existing technology is electricity.

Something who's ubiquity we take for granted which permeate's almost every aspect of our world.

summative assessment seen as harmful.

I went on a visit to a school open day for parents of prospective students, and I took the oppertunity to do a survey of the teachers present. I was asking about IT use in the main, but I asked the religion teacher how you teach religion. Well this obviously depends on the course, and as she said she is not looking to convert, but to inform and there is a lot of comparative religion material in the course. However one remark struck me: She said the students were generally fine and interested in first and second year, but when they got to third year and realized that there was an exam at the end of it, teaching them became more up-hill work.

Its been a while

I hope to re-engage here. This has always been a personal space for my own musings, and the fact that it is publicly viewable, is a property which I use to force me to exert some editorial control as opposed to just random typing. I think writing about educational matters is a useful discipline, as is a significant part of reflective self directed CPD.
To anyone who has followed me in the past, expect more here, I'm working on a couple of posts which will rollout when they're ready.

Thoughts from Thinking and speech (Vygotsky)

I'm returning to thinking about education as a practice, and this has involved me in reading more theory over the past couple of weeks.  To take a break this evening, I decided to watch a video from Nicolai Veresov on Vygotsky. In watching it, I was drawing an analogy between speech and teaching, and came up with the following thoughts.

Teaching is a means of social interaction. We lose the sense of it if we break it down into its component parts.

Teaching combines the function of social interaction and the function of thinking.

View Teaching as a unity of thinking and communication.

The social world is the engine of development. Other factors control the rate, but are not the engine. (pure Vygotsky here)

Adult teaching is a function of society. If you want to truly understand it, you have to see how its assumptions developed in the child. That is, consider the educational histories of both the students, and the teachers.

Marshal McLuhan and hot and cold media and other thoughts for eLearning

Marshal McLuhan and hot and cold media.

In building a blended eLearning resource for online learning, it behoves us (you don’t get to use that word often) to take a look over our shoulders at Marshal McLuhan’s work on media in the sixties.

Radio vs. TV. We find that students are often more inclined to download course materials as an Mp3 audio filer rather than as a MP4 Video.

McLuhan had this taped. Radio is a cool medium. It does not require our total attention. For eLearning this may not seem like something we want but we have to look beyond the immediate. The Mp3 file may be listened to multiple times as the student drives to work, goes for a work out in the Jim etc.

Video on the other hand is a hot medium it requires total engagement and its difficult to combine absorbing this material with driving a car for instance.

So use video sparingly. Use it as a welcome to the course maybe and to demonstrate specific content for which it is the most compact way of presenting it.

Present it in small chunks, and if possible allow the user to alter the speed of playback.

The next piece of McLuhan is the idea that the medium is the message. And associated with it, the idea that the content of any medium is always another medium. This one plays into our practice in two ways; the first is to be aware of what we are saying by our choice of delivery system. If I prescribe the exact sequence in which you will peruse the learning content I am giving you, I may think that I am making informed and sensible choices on your behalf as to what is the most appropriate sequence in which to consume my content. You on the other hand may see it as an arrogant and inflexible set of assumptions as to just how ignorant the learner is.

The second  concept here, that the content of any medium is another medium is more straightforward to apply. If my video is just me speaking, then the video content is audio. If my audio is me reading from a prepared script, then the content is text.

Maybe we are in a position to cut out some of the media middlemen.

Thoughts after watching “a day made of glass” and the “Microsoft Productivity future vision” pieces.

Thoughts after watching “a day made of glass” and the “Microsoft Productivity future vision” pieces.

Education is about building internalized models of reality. These models are then used to inform our actions with respect to possible futures.

In the corning model, a model of light mixing was played with in a very high tech way on a light table. This example was presumably chosen as content which would be immediately clear when you saw it on the video clip.  The learning involved a group collaborating to explore a cool educational artefact. As a model of collaborative and group education it was a nice example.  The park scene showing the dinosaurs was another easy pick up with its reference to the movie Jurassic park.  

The Microsoft educational vision was weaker. Although the whole tone of the video was corporate, the educational segment showed drill and kill math teaching with 20 year old graphics of a Bear.  The use of high tech to avoid the effort of opening the fridge door… How cool/pointless was that?

Education is a conversation. It is a communication. It is articulated structured and may contain Satori moments, but at the end of the day it’s a two way flow of information.  The corning vision showed an enriched channel with more interactions going in more directions.

My vote Corning: +1, Microsoft: try again.

Moocing into the future

There has been a lot of comment about MOOC providers such as Udacity or Coursera being a disruptive technology in the filed of higher ed, and this is hard to argue with, although the process is likely to be more protracted than many commentators think.  However just because change is taking place, it doesn't mean that we have to be antediluvian about it.
I know this is a weak analogy, but I'm sure that when the motor car came out there were some people complaining about the effect the innovation was going to have on the lives of ostlers, grooms and dung collectors.
We don't know where new online and automated forms of education will take us.  What we do know, is that they have the potential to take us someplace very far from the current state of affairs.  This potential may involve some creative destruction, and toppling of sacred cows, but it is up to us to steer it towards the creative the innovative and the good, and away from the banal and the mass market.

A short thought on MOOCS

There are a number of different versions of MOOC and I would suggest that the versions promoted by George Siemens, and Stephen Downs et al, are the most interesting. I have taken part in some of these, and they are a loo sly organized scrum of people generating and discussing content around some core theme. Very much a collaborative and distributed effort which at times has much of the characteristics of an extended research conference.
The very recent development of Massive enrollment courses from such sites as Alison, Coursera and Udacity, are somewhat different, and strike me as being a somewhat transitional use of technology, like early plastics trying to look like more traditional materials and failing.
I have taken part in courses from all of the above. If one views these courses in the traditional manner, then the ability to respond to students is extremely limited, they scale very poorly, and the fall out rates are horrendous. However if you view them as something more akin to an encyclopedia, or revision notes for a topic, then the idea of casual usage of the material, without taking all the tests or listening to all the lectures becomes more understandable, and provides a new model for learning.
When I taught physics at university level, I used to tell my students that there was a very similar course available on line through MIT's open course-ware initiative. I said that if there were topics where my students found my presentation confusing that I recommended Prof Lewin's treatment as an alternative. As he was paid more than me, and had been teaching the material for longer, it could potentially be clearer. My students responded well to this.
So I think we should view these courses as another tool to use in our teaching, playing a role similar to that of an academic library where you went to look for other books on a subject to either explore it further or get a better explanation. All contributing to the ultimate task of dispelling ignorance.

Thoughts on teaching math with calculators.

I teach some elementary math’s to adult learners, and I am mandated to teach them how to use a calculator. 

When I produce the calculators about ½ way through the course I ask them what a calculator is for. I then explain that it is for when the calculations get to big to do by hand, and we then move on to calculate some really big numbers, like the amount of sand on the local beach, the number of snowflakes that fell in the last big storm, the number of breaths they have taken so far in their lives etc.  I use the calculator to introduce the idea of estimating and approximating, and we then try to make an estimate of how good our answers are.  The idea is that, now you have the skills, the calculator opens up a whole new vista of doable problems.

That’s for my mainstream students. For some of the students, maybe two in a hundred, I teach more specific calculator use, when I have decided that these are students who don’t have the ability at this time of mastering the skills of hand calculation. This may be an innate deficit, or it may be the product of years of confrontational math education resulting in the student being terrified by the prospect of looking at numbers in any constructive manner whatsoever.

These are the students for whom the calculator will become the mathematical equivalent of dragon dictate, and just as dragon dictate can open up the world of generating texts to individuals who would be unable to write or use a keyboard, I see the calculator as giving access to mathematical results to people who would otherwise be incapable of getting there.

Finally to mention some specific skills. I teach adding fractions by cross multiplying and reducing the result. I explain that there are other methods which they may already know, but that I am avoiding them because they don’t always work, and we don’t have time to go through all the special cases.

I do this, not because I’m short on time, but because often I find students abandoned their attempts to understand fractions at the LCD/LCM stage. Fractions are key, and I explain that these operations are important because if they come to algebra many of the same methods will work. 

I spend almost a full class on the concept of dividing by a half.

I teach long multiplication, again stressing that this is a method which will go forward into algebra, and I teach long division in the same way for the same reasons.

The key in all of this is that I am teaching skills which will continue to be of value, both in general life, and in the students further math’s careers, should they care to pursue them.

Instructional design and Tacit learning

These are terms which hover around the borderland between education and training.  Instructional design came originally from a military need to train lots of people quickly to a uniform standard in the Second World War, and a desire to take educational theory and use it to produce a effective framework for people trained to a specific set of competences. Gagne is a key name here. Within that context, Tacit knowledge is the collection of nonverbal learning that SME’s (Subject matter experts) gain from experience which can be difficult to encode into a formal description of the specified learning outcomes of a course.

Once we step out of that context it becomes more difficult. I see education proceeding from very well defined learning goals, such the ability to tie a shoe lace, or to add decimal numbers up two places of decimals at primary level, through less concrete goals at secondary level such as “Students will understand the causes of the civil war”.  As education proceeds through third level, we encounter the sort of concepts embodied in the idea of a liberal education, or Newman’s idea of a university. Here packing what is being learnt into a Specified Learning Objectives framework becomes less appropriate.

I am referring by analogy to this type of learning also as Tacit learning, and in particular I am  thinking of the kind of skills which we develop as a professional, which are open ended, and don’t have set limits.  Good teachers, I think, don’t stop learning when they are qualified, and learn from a diverse set of sources including their students, and their practice.

This is an area of learning for which I can find few instructional design models, and which I am interested in exploring further.

Searching for assesment

This piece is a result of reading this blog post http://opencontent.org/blog/archives/2237

There has recently been a considerable amount of controversy about employers asking for job candidates Facebook credentials in order to find out what sort of people they were, and what sort of things they got up to. 

The critical issue seems to revolve about the distinction between a users public and private life.

In practice we may be judged by the company we keep.  This issue is particularly apt because Facebook is primarily a social network, and unless I am a policeman, and my friends are mostly journalists, it’s not totally clear that this information is relevant to an employer.

Be that as it may, David Wiley of Bringham Young University  has come up with another scary evaluation metric.  How about letting your employer, or your prospective college, see your search history?  Using “Big Data” tools, this could give either power broker a deep insight into whom they are potentially entering into a long term relationship with.

I think, with a moment’s thought, that this one too comes into the too much information category, and would fail European data protection legislation on the basis of informed consent and need to know grounds.  It would also leave a potential employer open to charges of discrimination, by providing them with information on which to make a discriminatory decision.

The problem again comes in because of the umbrella nature of such a request. It grabs everything, and a lot of it is outside the scope of any possible legitimate use of the data.

Now how about if I am asking my students to do research?  Suppose I’m asking them to document their web journey in search of enlightenment on the subject using membranes to provide high altitude diesel trains with oxygen enriched air for combustion.

In this case, it could be legitimate, and could provide a valuable anti plagiarism tool.  It shows how long they were on task, It shows how efficiently they used the web, and it shows how successful they were while they were doing it.

I could potentially have a number of target websites which I expected them to find, and I could evaluate them on a richer metric which could combine time on task with success in locating information, and a subjective assessment of their search methodology. 

Because the search could be seen as coursework, and could be limited to a defined time period, we could avoid the privacy issues, which down check the more general histories covered above.   As for an employer, I could ask the candidate in advance to provide a transcript of an hour spent researching the company prior to the interview.  This could be an effective test of their skills and prior knowledge in the field. 

Teachers are peasants!

Introducing innovation to teachers is difficult, because they operate in a subsistence manner. i.e. They operate at the limit of their resources and are required to get a crop in each and every time.Hence the peasant analogy.  This makes them disinclined to take risks with sweeping changes which could make them loose a years crop, but open to small labor saving incremental changes.

Entropy as a measure of class learning

I has just occurred to me that as teachers we work a piece of un-physical magic as we lead or guide a class through a course.
My reasoning goes as follows. At the start of the class, all the students are unknowns. We don't know how well they learn, how hard they are going to work, or what their ability is.
So if we define the entropy of the system in terms of all the possible outcomes, any student could potentially end up at any point in our class and the entropy is high.
in formula S = k Log(n)
where n is the number of possible states of the system, that is the number of different potential outcomes.
If I have 30 students in the class, and I mark out of 100, then this comes to
S = k 30 Log (100) = 60 k.
After the first mid term,  when 30% of the marks have been decided,
 S = k 30 Log (70) = 55.4 k.
After the second midterm with only 40% of the marks to be decided by the final,
S = k 30 Log (40) = 48.1 k.
After the final, when all outcomes are clear,
S = 0.
As the semester (term) goes on the probable trajectory of each student becomes more constrained. with the system eventually resolving into one final configuration, the final marks, and the uncertainty about the progress of any given student has reduced to zero.  In thermodynamic terms, we have cooled the system by removing the entropy and condensed the educational gas which the students represented, into a set of distinct phases, one for each possible student outcome (mark)..

We have reduced the entropy of the system Can I say cool?

Learning is one dimensional

Not all of it. But I have encountered a problem and its made me think of learning in a new way. I wanted to use Moodle to asses student learning styles, this would involve asking the questions from a test bank, and then rating their answers along a number of different dimensions of learning preference.
I have not solved this problem.  We normally assign students to a point along a one dimensional continuum which runs from didn't understand a thing I taught them in class to has complete understanding of everything covered.
Our aspiration as teachers is to move student along that scale, with the hope that they will end up at the I know everything end of it.
We are neglecting the possibility that their learning is moving in several different dimensions at once.  They may be failing my maths course, but becoming more reflective and capable persons through knowing their limitations.
They may be learning a lot about the interpretation of data and the application of geometry, but failing to pick up on the abstractions of algebra.
In most subjects, if I think about it, I can find several different strands or dimensions of learning, but we insist that at all times is all part of one bag which we judge solely by how full it is.

Maybe its time to make our subjects more multidimensional.

An online course with a good pedigree CS101

Last fall, I did an free online course in AI last autumn with Professor Sebastion Thrun of Stanford.

The course was very good, and about 22,000 student completed it.

Professor Thrun has now joined a new startup company, Udacity, who are offering two courses to start out. One involves learning how to program a robotic car (Proffesor Thrun was part of the team who won $2,000000 for building aan autonomous, self driving car.)  and the second is about building a search engine.

I think either of these course would be a great project for a TY student, or would be great for someone looking to increse or brush up their computer knowlege.

I have enrolled in the build a search engine course, which aims to teach complete beginners computer programing in seven weeks.

I'm doing so to brush up my programming skills, posibly learn a new language and to learn about programming to interface to the web.

I'm passing this information on as I fell it may be of interest to other members of the teaching comunity.

The web site is here:

http://www.udacity.com/cs#101

and I would invite you to at least view the introductory video, which includes a recomendation by one of the founders of Google.