omenti

moving from noun to verb

About a week ago Apple announced its latest Watch.  While it isn't a major leap, Horace and Om wrote great pieces about something more significant - namely what it does.¹  The first few iterations were a bit muddled - now it has become a wearable assistive technology.  That's significant.

 

Assistive technology invokes images of wheelchairs, braces, crutches and so on..  adaptations to make life easier for those people with physical problems.  It's a mistake to think that way.  All of us use these technologies  (I'll ignore the distinction between assistive and adaptive) ..  utensils, chairs, shoes, door knobs, screw drivers - all of these count.  Many of us wear eye glasses - a technology that compensates for a deficiency that would be a serious handicap if not corrected.  Some people have artificial joints and pacemakers - implantable assistive technologies that are always worn.   It begs the question of who gets to be called normal and who makes the decision.  I've written about that, so back to the Apple Watch.

 

At the beginning of the nineteenth century town clocks were an adaptive technology that gave people in emerging industrial cities a sense of common time to allow for synchronized work.  Less expensive models for the home became a necessity for workers who could afford them.  People who couldn't employed a variety of other techniques including "knocker-uppers" - people you would pay to make sure you got up in time.  Towards the end of the century the need for standards and a more global time synchronization emerged - railroads were the first big driver.

 

Railroad time is to the time of the future. The Sun is no longer to boss the job. People—all 55,000,000 of them—must eat, sleep and work as well as travel by railroad time. It is a revolt, a rebellion. The sun will be requested to rise and set by railroad time. The planets must, in the future, make their circuits by such timetables as railroad magnates arrange. People will have to marry by railroad time, and die by railroad time. Ministers will be required to preach by railroad time—banks will open and close by railroad time—in fact, the Railroad Convention has taken charge of the time business, and the people may as well set about adjusting their affairs in accordance with its decree… . We presume the sun, moon and stars will make an attempt to ignore the orders of the Railroad Convention, but they, too, will have to give in at last.
—Indianapolis Sentinel, 21 November 1883

 

Synchronizing clocks turns out to be an interesting technical problem. A Swiss patent clerk named Albert Einstein specialized in clock synchronization patents by day while thinking about time more deeply after hours.  

 

With atomic clocks - some orbiting in GPS satellites - we now have very accurate time synchronization available anywhere on Earth.  It's available in your smartphone and your Apple Watch with much better accuracy than the display requires .. very accurate time is much more central to communication and location determination. 

 

So just how important is a watch as something you wear to tell you the time - the watch as a simple noun?

 

There are still synchronizations for connecting with people, transport, broadcasts and so on.  Some of that requires less synchronization.  Some of the most creative work is poorly synchronized much of the time.  Television has become streaming video at your convenience and much of radio has turned into podcasts and streaming audio.  So why wear a thick $400 piece of hardware on your wrist?

 

You wear it because it is becoming a verb.  It's stuffed full of sensors and watches you.  A growing list of health, exercise and motivation related functions.  Complex computations turn thousands of dollar of tech in a doctor's office or researcher's lab into a few one dollar sensors and software.  And all of this can be networked.  I'll mention Happy Bob - a Type 1 Diabetes app by one of you (Jutta Haaramo) that is making the lives of thousands of diabetics easier.   This is only the beginning.    The Watch overlaps a bit with the iPhone, but they mostly fill very different functions - wearable is a big deal.  It's important to note the Watch still has a lot of headroom for growth. 

 

We humans have been expanding and augmenting the basic human body for over ten thousand years and the process is accelerating.   It's exciting, but we should be thinking more deeply about what it means as well as who is normal and who decides.

__________

¹ Oxygen saturation wasn't difficult to add - it exists elsewhere.  What is important is the watch can keep a record.  A General Practitioner friend is currently testing one. If it takes good readings he's going to recommend it to patients who are in the at risk categories for Covid-19 - cheap insurance. 

That said I think the major improvement was the addition of an ultra-wideband chip. I wouldn't buy a device just for that as this is preliminary work for something that could be very big, but it's good to know new iPhones and Apple Watches have the capability now.

eliza the next generation

At the beginning of each year I take a two week mini-sabbatical to explore something outside of my field, but interesting nontheless.  Around 1985 it was the computer language Lisp (Common Lisp because one of the machines in research had it and a tutorial).  The best way to learn a computer language is to give yourself a task.  I wanted to build a chatbot - specifically recreate ELIZA.  It wasn't particularly difficult.  There was a very small (like a half page of code) main module and scripts of responses.¹  I only had two scripts - both were simple, but they worked. 

 

ELIZA was created by Joesph Weizenbaum at MIT in the 60s to test human computer interactions.  The most famous version was DOCTOR where ELIZA was a psychiatrist.  It was primitive by the time I tried to recreate the basics, but it blew people away in the day.  At least one magazine predicted this was the end of any number of jobs.   It also made people think and rethink what the Turing test was. Since then there have been a variety of chatbots of varying degrees of sophistication.  You've used more than a few interaction with companies over the telephone or via the Internet. 

 

Times are changing.  I'm reacting to chats I've had with friends who do 'AI' (such a loose term!) at Stevens, Columbia, Google and Facebook as well as a recent paper that shook up some people.  These conversations lead me to a prediction:

You should learn about Large Language Models and be wary - very wary.

LLMs are statistical models of words that we might use.  A very large corpus of writing - say Wikipedia, blogs, everything posted on social media, some public domain work - and even computer code (GitHub for example) - is used to find strings of words and statistical probabilities of what comes next.  These are computationally expensive to build so expect them from the largest tech companies and governments. Not many people have this expertise and they're mostly academics. But once you have one, the task of building a front end to interact with people is cheap in terms of talent, time and compute power. 

 

They have no intelligence and can't make reliable predictions (although I suspect horoscopes are very likely:-).  Early work suggests they can be appear authoritative even though they're mere, as Emily Bender suggests, mere stochastic parrots.  I suspect a lot of people will swoon and some real damage can be done before the right niches are found.

 

Here are a few I can think of (some suggested by folks I know)

translation in real time  Not the quality you'd use at the UN or for contracts, but better than high school French or German.

transcription  Human transcription is very expensive.  For most talks and meetings you don't need the accuracy of a competent human. 

spam  This class of tool should be able to write highly targeted spam that can fool a lot of people.

national destabilization  We've seen what it's like with human authors and tv and social media platforms.  Expect certain players to exploit these tools when they're available.

writing  Book reports for kids to journal articles.  They can be very misleading.  In areas where the corpus isn't deep expect to see serious plagiarism.

search  This one really scares me.  Authoritative answers that can be very wrong. Of course people will sign away any legal rights they have before being allowed to use these. They might serve as a writing tool to rephrase something poorly written. Think of it as a next level spelling/grammar checker.  

there have to be many we haven't thought about!

 

It will happen in less than five years.. maybe even next year given what I've seen so far. My advice to to think about implications.  I worry too many bad ideas are the result of running fast and breaking things.

And since I mentioned plagiarism:

 

__________

¹ Here's the pseudocode:

function Generator(user sentence) returns response
    Let w be the word in sentence that has the highest keyword rank
    if w exists
        Let r be the highest ranked rule for w that matches sentence
    response ← Apply the transform in r to sentence
    if w = 'my'
        future ← Apply a transformation from the ‘memory’ rule list to sentence
        Push future onto the memory queue
    else (no keyword applies)
        Either
               response ← Apply the transform for the NONE keyword to sentence
        Or
                response ← Pop the oldest response from the memory queue
    Return response

 

scribbler moon

Margaret Atwood

David Mitchell

Elif Shafak

Han Kang

Karl Ove Knausgaard

Ocean Vuong

 

Each of these authors has taken a walk through Nordmarka Forest in Norway to place a their manuscript in a special room in the Deichman Library.  Ninety four more will follow.  The authors are specially chosen for their imagination and the stories are to based on the themes of imagination and time.  No one other than the authors will see their words until 2114.   In that year someone who doesn't exist now will take each of the 100 manuscripts from their containers and prepare them for printing on paper made from 1,000 one hundred year old Norwegian spruces that were planted just before the Margret Atwood took the first walk through the forest. 

 

The Future Library project was created to stir the imagination and give hope.  A century is a good time period - just outside an average human lifespan, but not so long as to be unimaginable. What will be around? Will people care?  

Our greatest responsibility is to be good ancestors - Jonas Salk

 

We live with a number of legacies - a written language, medicine, science, religion, weapons, slavery, pollution. ... some good and some bad. Mostly we live in the present and don't think much about the leverage we have in creating future legacies. Some legacies happen by default.  Humans are one of the few species with a sense of the future.  Projects like the Future Library spark wonder and perhaps give us hope that maybe we can make a difference.  Perhaps a trick is to think about time scales and what we might accomplish.

 

You can do something for yourself on a smaller scale. Here's one of mine:   My thesis advisor gave me his favorite slide rule when I received my Ph.D. It had been given to him by his advisor when he received his.  It became my favorite.  I used it regularly and wondered about the legacy of calculations that were used to build conjectures and hypothesis through their careers.  Every now and again I wondered what I'd do with it.  About twenty years ago a very unusual Summer student came along.  Art and computer science, she ended up at an animation company.  In slide rule had been used for a certain type of story telling..  small corners of the fundamentals of Nature.  I gave it to her some time ago so I could enjoy the link between the past and the future.  One type of story telling changing into another.  She says it's very special to her and gets use for calculations used in making films that tell stories.  Wondering who she gives it to and what path and stories it might inspire gives me pleasure and hope.

 

In my work I have to think of a variety of time scales.  Some events at the subatomic level take place in a under a billionth of a billionth of a second.  At the scale of the Universe some things have been around for billions of years and will be around much much longer.  These times so foreign as to be abstract.  A few weeks ago one of you sent a note from the Grand Canyon that was filled with wonder inspired by that slow geological process that created it.  Wonderful experiences to stir the imagination.  We should seek out these experiences, but I find staying in human time scales and thinking about and building legacies lets me live in a slightly better future and gives me hope.    Set your mind some distance in the future and think of something good you'd like to see there or something bad you'd like to change.  With a target maybe you can imagine a path and help work towards it.  It gives us a way to not discount future people and their world.

 

Scribbler Moon?  That's the title of Margaret Atwood's book you won't read, but maybe your great great grandkid will..