I'd been putting off watching the Jeopardy Watson video as I assumed it would be nothing more than your typical orb of visualised intelligence speaking in a friendly voice and putting humans to shame. Turns out I was partially correct. It was orb like, and it won, but the method of visualising a super computer was much more interesting than I thought. One of the issues in making what is nothing more than a series of servers, hard drives and some very clever algorithms is in making this approachable to humans. One can imagine the remake of 2001 using something similiar to Watsons impression of sentience as a stand in for the revamped HAL. It's the super friendly voice that does it. Makes it approachable, believable, yet without attempting to cross the vocal uncanny valley.
I've never really cared for Josh Davis. I didn't see the point of Praystation when I came across it in undergrad; his more recent generative works seem to be just a series of patterns, albeit created generatively. But I'm damn impressed with the approach that he took with making Watson an approachable and non threatening entity. I'm going to presume that this was created in Processing, mostly because it's all I can see at the moment, but I don't know of any other way that this could be made. What with the swarm dynamics and all.
I'm really interested in the many different personality traits they've given to it, as well as all the branching decisions that entail all of the different traits. I'm also rather pleased that they decided to go with the the IBM Smarter Planet logo as the face of Watson. Based on the video, it seems fairly evident that this decision came form on high, the stern IBM guy would have made his point in one of the meetings, and I'm happy that they got to pull rank. Note, I'm making this up completely based on the part in the video at 2:07. The use of a logo, or a screen based representation as opposed to something more robot like helps in conveying this as an intelligence far different than what we've come to expect.
It's one or two steps removed from the relational logos I described in my thesis, but close enough that it remains closely linked to their overall branding guidelines. I'm not sure something this detailed could work as a logo, and why would you bother? The Smarter City logo is clever enough (I had to work real hard not to dovetail a smart pun in there) and this is a rather charming extension of it. In my thesis I looked at three different categories: Contingent, Networked and Generative. This is firmly placed in the generative pile, but it goes one or two steps further than what I've seen before, and not just because we're dealing with a near sentient AI. Due to the 26 personality traits that they've given it, the level of complexity far outreaches any other possible generative logos.
One of the familiar tropes here is the use of the pulsing orb as a means of signifying the speaking entity. Not the most creative of decisions, but a useful one at that. At least we know that this will continue to be used in movies. I can't recall any at the moment that use this, but they have to be out there. Maybe HAL? This works as a nice and comfortable shorthand illustrating voice waves, so we're probably going to stick with it for some time. What is rather neat about this is how the colour of the voice relates to the emotion that Watson feels as he speaks, or as he responds to incoming stimuli such as the presenter.
What I'm most fascinated by is the use of swarm and particle dynamics as a means of illustrating sentience, thought and emotion. And not because I just learnt how to make it recently, although that does help. Davis mentions that he created a swarm, with one leader and the rest follow that swarms decisions. I'm not sure why he went with this approach, but it seems to work. It creates a sense of urgency in the movement, as well as keeping the whole thing in one cohesive group, while allowing for a controlled amount of variance. It's an interesting choice, using a swarm to indicate sentience, and one that relates somewhat to how the brain works. The human brain is not a monolithic individual, but composed of a multitude of different decisions, all working somewhat together to give the appearance of the singular. The same thing is clearly at work in the way Watson searches for the correct answer. It shows what I'm presuming to be a multitude of differing and competing algorithms all trying to evaluate their own decision and then seeing if they're confident about their own choice. This swarm is seen as unintelligent bots racing through the various possibilities contained within the hard drives or the internet. It's a clever choice, as it seems an entirely appropriate decision.
Despite this not exactly falling into the realm of logo design, I'd go so far as to place this in an entirely new category. It's partially generative, but only in so much that there are generative rules that dictate the animated form of the object. But this goes further than form as it begins to move into the realm of behaviour. There's a whole host of questions that arise once this starts to become, if not the norm, then at least unsurprising. Will we start to see this used on a more casual basis? As fractionally sentient branding marks begin standing in for a whole company, how will they create personalities that are able to represent the core brand values of said company? Will this then become part of the repertoire of branding? How will we, as users, consumers and citizens react to this new form of branding? What will be the role of the designer, will this open up a whole new host of design skills that weren't possible before? What will this new field be called? Personality Designers? Brand Sentience Managers? In writing about emoticomp, Ben Bashford has outlined some of the skillsets that might be needed for this new field:
What’s clear, and it’s been said before, is that there’s an opening for a new type of designer. Someone that understands interaction design, product design and can add character to things through behaviour. A light touch. Very subtle in order to make them believable - without them being too ridiculous.
He's talking about products and devices - tangible physical things - but there's a similar thread at work here. What if they start to piss us off, in ways that didn't happen before? That we decide, now that the brand has taken on a new facet, that we don't really like him. Will it be a female or a male? What if there ends up being more male brand AIs than female? That we really like the products they make, but can't stand the representation. I'm not referring to mascots, but that comes into it somewhat, but more abstract representations. We can't always go with the softly spoken somewhat british pleasant voice; there'll be children voices, elderly voices, accented voices. This scenario will only be possible for specific types of companies, but they won't need to be large multinationals, as the cost and scale of fractional AI will begin to be affordable for medium range companies.
What's fractional AI? Ask Matt Webb, he coined it, and it's a bloody useful term to describe the what I'm talking about when describing partially sentient brand ambassadors. It's not AI in the monolithic sense that Watson or HAL employ, but how small bits and pieces of AI can be used as minor interaction points. It's an hack, a cheap trick in creating believable sentience for a very small set of design and user needs. His How fractional AI can be used to make nicer things talk is a rather fascinating introduction to the topic but Botworld: Designing for the new world of domestic AI talk is a very considered and thoughtful take on how it might play out.
Back to the prognostication. What if the logo crashes? There's a bug and it starts to behave in a manner far removed from it's designated personality traits? When do we get the first scandal of a logo being improper? The history of Sci Fi is littered with robots and AI gone wrong, but these brand marks are limited to a very small subset of personalities. Yet it's entirely plausible that one of them will be unintentionally mean to me. Having your feelings hurt by a domesticated robot gizmo is fine (to an extent), it's limited to a particular and individual object. But being told I'm stupid, or fat by a brand, even by no fault except the algorithm will be a severe problem. Made even worse by these non human intelligences trying to remedy it.
The domesticated robot should be subservient to the human, both in it's job and in it's intelligence, but the artificial brand ambassador of the future should be a few clicks above that in terms of representation and perceived intelligence. Matt Jones refers to this as BASAP (Be As Smart As A Puppy), but this falls apart when you move away from small robotic companions and helpers toward representations of brand values that don't mesh with this need for emotional simplicity. When brands require sophistication, and use screen based representations we'll need a new model on which to base it.
The design behaviours employed with Watson rely on colour, position and speed of both the swarm and the orb. These all work rather well, especially in making him become approachable and non threatening, but there'll be a time when we need to move away from simple particle systems. Something tells me that neonatal representations of the human face won't seem appropriate. There's time yet before this becomes a reality, hell, there's time yet before this becomes remotely plausible. But we'll need to start thinking about ways in which generative forms can move up the levels of complexity while still operating in the realm of design.
Comments