CIO Folder: We need to talk… to our robots
17 October 2016 | 0
If this then that, if that then this plus other things. Or whatever. As we all know, that’s how computing works. It’s all relentlessly logical. Programmers and coder dojo kids understand this. They are right down at the base level of instructions, one tiny step at a time. Numeric only, although the system can be binary, octal, hexadecimal or decimal. Then you have multiple possible programming languages, some of them with invisible layers of pre-written code, like idioms in human speech.
“Robots with artificial intelligence will be programmed to tell the truth, we can assume, despite the speculations of science fiction. But that truth will by definition be self-serving because all the ‘intelligence’ knows is what it did and why it did it”
Most of us, however, certainly since the WWW and the beginnings of browser interfacing, have grown accustomed to using human languages. Our fortune on this island is that English is the dominant one and in fact the foundation language in which most applications are first composed, presumably because of the sheer dominance of US corporations in the global IT sector.
But of course the processor, the computer, the system cannot ‘understand’ as we understand. It can only process, according to the logic with which it is programmed. But that has all developed into extremely complex and sophisticated sets of capabilities, still based on those tiny units of instruction. Robotics and machine learning, machine vision, growing degrees of autonomy in vehicles and drones, are all strides forward at a rapid pace into a new world of intelligent and connected devices.
The next step, according to the established mythology of robotics, is artificial intelligence. Because AI means different things according to different lines of research, we can’t actually be sure what we mean by that. Does it mean replication of superior human intelligence, needing only the verisimilitude of artificial bodes and skin and features to pass for another human? That is the movie version of the vision. But Hal is another and arguably more likely or plausible concept. In 2001: A Space Odyssey the ship’s controlling computer communicates with human Dave in smoothly robotic speech but is really just a piece of hardware and built-in logic like any advanced computer.
We are running a little late on Arthur C Clarke’s vision, but HAL 9000 is in sight. Hal, the crew’s pal (for a while) in the film is named for a class of Heuristically programmed ALgorithmic computers. Heuristic is the programming trick to get closer to real life human-type decisions, because it allows for degrees of imprecision or approximation and making or choosing the best choices from possibilities that are not always optimal. A classic computing solution will do nothing if the choices do not fit defined parameters. Most of us remember programs that simply went into an endless loop when puzzled.
In our real life, when robotic controls are in place, self-driving vehicles are a good example, we want decisions to be made in real time or as close as possible. We want to trust that the computer intelligence, like humans, will choose the least bad of a set of undesirable options. Very seldom will that mean doing nothing. The upside is that the speed of decision making can be much faster than humans and situations (again, think of self-driving vehicles in danger) can change in seconds so new decisions will be required. Being unemotional and probably sturdier than our skin and bone, it is to be hoped that the robotic ‘drivers’ will continue to perform until it is impossible. So on that basis alone they would generally be safer than humans in extreme situations. Provided they make the best decisions. Not the ‘correct’ ones.
So far so good. What if there were a fatality on the road in that self-driving vehicle? The inquest would obviously need evidence from the driving system. How would that evidence be given? Would it be something like a computer log, explained by an expert? Or would the system be able to explain its logic and decisions in machine speech? A robotic system that could explain itself in free form human speech like HAL would be tremendously impressive. Ignoring the question of that ever being possible, would the fact of that verbal evidence — certainly involving answering questions spontaneously — be more convincing? Perhaps in fact it would have a spurious authority precisely because of the synthetic speech fluency.
Robots with artificial intelligence will be programmed to tell the truth, we can assume, despite the speculations of science fiction. But that truth will by definition be self-serving because all the ‘intelligence’ knows is what it did and why it did it. On the other hand, if the systems logs are what is produced in evidence, interpreted by a human, how persuasive would that be to our hypothetical future inquest jury? In any event, since a finding of dangerous driving, however unlikely, cannot result in prosecution such affairs will be in the realms of insurance law and claims against the systems or vehicle suppliers. We can immediately recognise the threat of an even murkier world of competitive legal truths than we have already.
Whether our robots are humanoid or vehicular or simply embedded, is likely to depend almost exclusively on the applications and markets. Science fiction may prefer to visualise beautiful women, armoured hunks and ultra-streamlined craft with go-faster stripes as well as the pocket nuclear power plant. But in real life it is a reasonable bet that most robots will be embedded in systems and services, dedicated to specific ranges of function. There is certainly room for mobile robots, from advanced Roombas to auto-butlers to maintenance technicians. We will almost certainly overcome our sentimental visions of two-legged automatons in favour of stable bases on tracks or roller wheels, multiple interchangeable arms or instruments and long life batteries.
The human-like interaction with us, the masters and users, is in many respects a separate line of development. It is and will be based on synthetic human language, which we are seeing proliferate and advance rapidly in Siri and Cortana, Google Assistant, multiple phone call handling systems and machine interfaces for user instructions. The design and construction of ‘personalities’ for these virtual assistants is certainly going to be a major line of development, perhaps including in the near future the freedom for a user/purchaser to compile a unique personality profile for a general purpose assistant.
It is more than probably that the AI engine for most of them will be virtual and in the cloud. That will not only add power and regular upgrades but make the assistant transferable to new physical devices and additional or new services. Tilly or Minty or whatever you called your first soft toy may now be your virtual assistant and companion for life, living with you into old age and remembering the things you cannot.
This suggests a future of augmented living, supported by a personal ecosystem of devices from transportation to cleaning to health monitoring to recreation. But that is just a bunch of gadgets, in the end of the day. Much more interesting is the possibility that your Assistant Alicia (homage to Ex Machina and its star) will be the custodian of your privacy and data and the intermediary or interface with all of the services you use.
She will learn your taste in coffee and changing video taste but also your health and how to take care of it. Your online transactions will be managed by her, certainly every aspect of your cyber-security. She will probably look after your money also, budgets if you need them, as well as doing the necessary checks into the background of your next (cyber) date. Actually Tinder may be replaced by Alicia talking to Tilly and fixing their owners up.
But don’t get too hung up on that mental picture of Alicia Vikander. Your future Alicia is more likely to be a football sized unit or a mobile smart omni-tool the size of an old StarTac or even a micro-SIM that you can lose and ruin your life. Unless she is properly backed-up. Which Alicia will ensure.