Exhibition at the Weston Library
On Saturday morning I met up with a friend at the Bodleian’s Weston Library in Oxford. On entering the building, we paid a quick visit to a small exhibition, Imagining AI, curated by Professor Ursula Martin, who has considerable knowledge of the history of computing, and has written and presented about Ada Lovelace’s significant contribution.
This exhibition had a strong literary theme as it drew mainly on the Bodleian’s special collections, putting on display rare books, manuscripts and other documents. The introductory panel invited us to “explore the boundary between humans and machines”. I found this an interesting, even provocative pitch, setting humans and machines on opposite sides, like adversaries. This was then extrapolated. To paraphrase: will machines foster some fantastic utopia or nightmarish dystopia and take over the world? Lest we dwell in happy reverie, the selections, just to the right of that panel included extracts from Mary Shelley’s Frankenstein.
With the (surely, generally accepted) observation that machines are more than mere “number crunchers”, the exhibition developed the theme of computation, highlighting the work of Charles Babbage and Ada Lovelace on the Difference Engine (constructed) and Analytical Engine (conceived, never built). An underlying theme, supported by a quote from Lovelace, was human mimicry, especially of the brain’s thought processes. W. Stanley Jevons aimed to demonstrate rationality by his “logical piano”, on loan from the History of Science Museum. More subtle forms of human mental activity, considered by some to be among the first AI programs, were expressed in the simplistic love poetry generator of Christopher Strackey (as re-implemented (or re-enacted?) in modern code) – more than a decade before the more well-known Eliza. These developments were explored and brought up to date in an associated event (which I didn’t attend), evidencing how the complexity has increased exponentially and has become prominent in the arts, where there’s the ‘Wow!’ factor – picked up by a local paper, which invited readers to meet Ai-Da, a superstar robot artist.
Thus, with all these bells and whistles, there is a sense of awe, even reference around these “transformative technologies”. But who or what is being transformed and in what ways? Is our well-being transformed?
The exhibition begs so many unanswered questions. I join the many who have claimed the existential threat is real and urgent, but what is less said and which I feel is just as imperative, if not moreso, is that in order to address and properly evaluate AI, we need to balance the discussions of machine ability with human ability; we need to remind ourselves and explore further what makes human life so remarkable in terms of originality and creativity. Furthermore, we need to reflect on the qualities of human sentience – to be and know – and how the presence of machines impacts this.
Without such a perspective, it’s too easy to fall into the trap of making (belated) comparisons under a narrow instrumentalist framework. Especially when millions of pounds are invested to demonstrate measurable benefits. As the sophistication increases, the accompanying language sometimes becomes hyperbolic and even idolatrous. Whilst computation elicits a myriad of possibilities to excite, we should not lose focus on the nature of the human mind, of its often-sublime qualities such as autonomy, reflection, intuition, originality, purpose and perseverance. All of these contribute to human agency and hence human flourishing and happiness.
Ada Lovelace was quoted expressing the power of these machines, but I think she herself might have given greater prominence – particularly given the exhibition’s title – to what she herself had expressed about imagination:
Imagination is the Discovering Faculty, pre-eminently … It is that which feels & discovers what is, the REAL which we see not, which exists not for our senses… Mathematical science shows what is. It is the language of unseen relations between things… Imagination too shows what is … Hence she is or should be especially cultivated by the truly Scientific, those who wish to enter into the worlds around us!
Journal entry for 5 Jan 1841, Lovelace Papers, Bodleian Library, Oxford University, 175, folio 199
It is the REAL, not the computational that Lovelace emphasizes that should be strived for.
A chatbot reactivates an expired SIM card – or does it?
On leaving the Weston, I popped next door to Blackwell’s to buy a couple of books. Only one was in stock so when I returned home I went to make a purchase online, using Paypal. Using the payment facility required two-factor authentication, which meant a text message to the SIM card in my trusty mini phone. But, on switching it on, it greeted me with the message that the card was non-operational. It turned out it had expired because I had not used any chargeable services during the past 3 months.
I logged onto the Vodafone site and proceeded to enter into ‘dialogue’ with its chatbot, an experience I had managed to avoid previously. Somewhat sceptical, I commenced the mission to reactivate the card with the help of TOBi, Vodafone’s ‘friendly digital assistant’, which has already been around for 5 years. It is mentioned that the service is powered by IBM Watson and LivePerson, which are well-established systems using advanced machine-learning.
Pedigree is all very well, but what about the user experience? My conversation proceeded, or sometimes stumbled along, as follows (note that I lost some formatting in my copy and paste, so there may be some errors and/or omissions):
[14:50]
TOBi: I’m TOBi – Vodafone’s
chatbot
Ask me anything, I’m here to help.
[14:51]
Paul: my number has expired, but I
would like it back
TOBi:
Let’s see if we can get your number reactivated.
First, I need to know what kind of customer you
are?
Pay monthly
Pay as you go
Paul: Pay as you go
TOBi: To keep your Pay as you go number active, you need to use it for at least one chargeable activity – like sending a text or topping up – every 180 days.
If you don’t top up or use any credit for 90 days, we’ll send you a text explaining that your account will expire in the next 90 days.
To avoid this, simply use your phone for a chargeable activity after you receive the 90-day reminder text. This will let you carry on as normal with your current number.
However, if you don’t use your phone within that 90-day period, your mobile number will be disconnected and will no longer work.
If your Pay as you go number has been deactivated due to inactivity and you’d like to reactivate it, please speak to an adviser.
Read more in our Pay as you go terms and conditions
<hyperlinked option> speak to an adviser
TOBi: Could you tell me your full name please?
Paul: <full name>
[14:52]
TOBi: To help save you
some time when speaking to an adviser, I’ll
complete some quick security checks with you
first.
For this security check, I’ll need to take your phone number to send you a security code by text. I’ll launch a secure pop-up for you to enter your number now.
<didn’t see any popup>
TOBi: Thanks, that’s the security part completed.
TOBi: Could you please tell me your phone number?
Paul: <entered the number associated with the expired SIM>
[14:53]
TOBi: hello
TOBi: I hope you are doing well today
TOBi: How can I help you today ?
Paul: i wish to reactivate my expired number
[14:54]
TOBi: I will try my best for you
but it is difficult but still I will try my
best
TOBi: Please help me with the sim serial number on the back side of the sim card
Paul: <removed case and battery, then SIM card, squinted and typed the 24 alphanumeric code>
[14:55]
TOBi: Thank you let me check and
try
<no screen activity>
[15:00]
TOBi: We have not heard from you
in quite some time. Seems like we have lost
connection. I’ll give it a minute more before I end
this chat session.
[15:01]
Paul: Hello, have you reactivated
the sim card?
TOBi: Just doing the process for you and it is almost 90% done
[15:02]
Paul: OK, thanks. I
shall wait.
[15:05]
TOBi: You will be very very good
to know tat I have raised the order and your number
will activated with in 2 to 24 hours
TOBi: I hope it is sorted for you all 🙂
Is there any thing else I can do for you please ?
[15:06]
TOBi: I am glad that have been
resolved 😀
Paul: Thank you for your help. Much appreciated. Goodbye.
TOBi: I would request you to please end this chat by clicking on the X button on the top of the chat
Conversational Quirks
TOBi maintained a chirpy and seemingly cheerful disposition. The issue was correctly identified at the outset and after 15 minutes, it seemed it had been properly dealt with; we got there in the end. However, I found the process somewhat precarious and it was a quite disorientating experience. I felt uncertainty in several places, but whereas with a human service assistant I’d be confident in checking the status, here I was worried that I’d be misunderstood and then be back to square one. Altogether, there were several hiatuses which made the conversation feel like an obstacle course:
- At 14:52, TOBi mentioned speaking to an advisor, but I wasn’t sure of the implication – whether it meant in the event of the bot not being able to handle the request at some stage down the line or whether I would be transferred imminently. There was no preparatory statement along the lines of, “a (human) advisor will deal with the query … “
- The absence of the ‘secure’ pop-up and then the validation confirmation without my consciously doing anything was a bit unsettling and surely something that needs to be checked by technical staff. And only after passing the security check, was I asked for my phone number, which I would have thought would have been part of the check
- Next (at 14:53), it was bizarre to be greeted afresh. At that point I had no idea about the status of my request, so I simply repeated it, echoing back the give terminology, “i wish to reactivate my expired number” (humans can learn too!)
- From (14:54), the style of English became quite different, so perhaps at this stage I was dealing with a human? But the instructions were clear enough and so by 14:55 I had provided all the information needed.
- But time was ticking by and these chat systems have quite strict timeouts, so at 15:00 I was threatened with a disconnection, with no reference to the status of my request. I wanted to check the status, but didn’t want to interrupt anything, so I just asked about the reactivation of the SIM and after a moment I got the response that it was nearly done. A few minutes later I got confirmation that it was done and I could conclude the chat.
And, as promised, the card was reactivated and working later that afternoon, which was a positive outcome (and a relief).
Observations
One of the disorientating aspects, possibly caused by my imagination, was my impression that in some places there was manual override that didn’t work well in concert with the basic chatbot workflow. Interleaving of processes requires some orchestration so that you don’t get disconnected whilst some useful process is happening … The chatbot seems to have been architected as essentially a sequential system, but this kind of troubleshooting is really a problem in parallel and distributed computing, especially if multiple agents (bot and humans) are involved.
I wonder to what extent the actions are being
carried out by the bot and what by humans. I
imagine that manual override is possible at any
stage (as with a self-driving car). It probably
occurred here as, surely, TOBi is not programmed to
write, “You will be very very good to know tat I
have raised the order and your number will
activated with in 2 to 24 hours”
(sic).
Or is that meant to be ultra-realistic human
interaction, complete with typos?
As to the problem of SIM card expiring, the timescales have shortened – I originally bought the phone with the number more than 20 years ago and don’t recall any expiry at the time. Eventually, it was 12 months of inactivity. Now it has been reduced to 6 months, but I only received one reminder, in the middle, i.e. with 90 days remaining. This is a long notice period, easy to forget. It would be better to have received another reminder with 30 days to go.
Even with optimised machine-learning and having my problem solved, I somehow felt a lack of completion. Human communication is actually much more sophisticated and subtle in its interactions, with a remarkable amount being non-verbal and intuited. I am conscious that with bots, humans need to fit around the ‘thought processes’ of the machines – I found myself trying to work out what terminology to use, and wasn’t sure of it.
What about the staff that TOBi has replaced? Do they really have more interesting work to do? Do they still have jobs? Or is the situation like high street banks, which have reduced in number and where the role of many staff has become far narrower? Are human chatbot assistants like those who hover around to help people use ATMs and deposit points, answer a few general questions, before directing customers online? I.e. to adjust to the ways of the machines … ?
The subject of artificial intelligence is of ongoing interest in my research pages on mind.
This page was published on 27 September 2022 and last updated on September 28, 2022 .