An MIT Technology Review article introduces the man responsible for the 30-year-old deep learning approach, explains what deep machine learning is, and questions whether deep learning may be the last significant innovation in the AI field. The article also touches on a potential way forward for developing AIs with qualities more analogous to the human brain’s functioning.
Programmers of social media apps and increasing range of other apps apply neuroscience to program your brain.
It’s common for brain functions to be described in terms of digital computing, but this metaphor does not hold up in brain research. Unlike computers, in which hardware and software are separate, organic brains’ structures embody memories and brain functions. Form and function are entangled.
Rather than finding brains to work like computers, we are beginning to design computers–artificial intelligence systems–to work more like brains.
An article at Wired.com considers the pros and cons of making the voice interactions of AI assistants more humanlike.
The assumption that more human-like speech from AIs is naturally better may prove as incorrect as the belief that the desktop metaphor was the best way to make humans more proficient in using computers. When designing the interfaces between humans and machines, should we minimize the demands placed on users to learn more about the system they’re interacting with? That seems to have been Alan Kay’s assumption when he designed the first desktop interface back in 1970.
Problems arise when the interaction metaphor diverges too far from the reality of how the underlying system is organized and works. In a personal example, someone dear to me grew up helping her mother–an office manager for several businesses. Dear one was thoroughly familiar with physical desktops, paper documents and forms, file folders, and filing cabinets. As I explained how to create, save, and retrieve information on a 1990 Mac, she quickly overcame her initial fear. “Oh, it’s just like in the real world!” (Chalk one for Alan Kay? Not so fast.) I knew better than to tell her the truth at that point. Dear one’s Mac honeymoon crashed a few days later when, to her horror and confusion, she discovered a file cabinet inside a folder. A few years later, there was another metaphor collapse when she clicked on a string of underlined text in a document and was forcibly and instantly transported to a strange destination.
Having come to terms with computers through the command-line interface, I found the desktop metaphor annoying and unnecessary. Hyperlinking, however–that’s another matter altogether–an innovation that multiplied the value I found in computing.
On the other end of the complexity spectrum would be machine-level code. There would be no general computing today if we all had to speak to computers in their own fundamental language of ones and zeros. That hasn’t stopped some hard-core computer geeks from advocating extreme positions on appropriate interaction modes, as reflected in this quote from a 1984 edition of InfoWorld:
“There isn’t any software! Only different internal states of hardware. It’s all hardware! It’s a shame programmers don’t grok that better.”
Interaction designers operate on the metaphor end of the spectrum by necessity. The human brain organizes concepts by semantic association. But sometimes a different metaphor makes all the difference. And sometimes, to be truly proficient when interacting with automation systems, we have to invest the effort to understand less simplistic metaphors.
The article referenced in the beginning of this post mentions that humans are manually coding “speech synthesis markup tags” to cause synthesized voices of AI systems to sound more natural. (Note that this creates an appearance that the AI understands the user’s intent and emotional state, though this more natural intelligence is illusory.) Intuitively, this sounds appropriate. The down side, as the article points out, is that colloquial AI speech limits human-machine interactions to the sort of vagueness inherent in informal speech. It also trains humans to be less articulate. The result may be interactions that fail to clearly communicate what either party actually means.
I suspect a colloquial mode could be more effective in certain kinds of interactions: when attempting to deceive a human into thinking she’s speaking with another human; virtual talk therapy; when translating from one language to another in situations where idioms, inflections, pauses, tonality, and other linguistic nuances affect meaning and emotion; etc.
In conclusion, operating systems, applications, and AIs are not humans. To improve our effectiveness in using more complex automation systems, we will have to meet them farther along the complexity continuum–still far from machine code, but at points of complexity that require much more of us as users.
New Scientist article: Applying the mathematical field of topology to brain science suggests gaps in densely connected brain regions serve essential cognitive functions. Newly discovered densely connected neural groups are characterized by a gap in the center, with one edge of the ring (cycle) being very thin. It’s speculated that this architecture evolved to enable the brain to better time and sequence the integration of information from different functional areas into a coherent pattern.
Aspects of the findings appear to support Edelman’s and Tononi’s (2000, p. 83) theory of neuronal group selection (TNGS, aka neural Darwinism).
Edelman, G.M. and Tononi, G. (2000). A Universe of Consciousness: How Matter Becomes Imagination. Basic Books.
Good discussion that covered a lot of ground. I took away that none of us have signed on to be early adopters of brain augmentations, but some expect development of body and brain augmentations to continue and accelerate. We also considered the idea of bio-engineered and medical paths to significant life-span, health, and cognitive capacity improvements. I appreciated the ethical and value questions (Why pursue any of this? What would/must one give up to become transhuman? Will the health and lifespan enhancements be equally available to all? What could be the downsides of extremely extended lives?) Also, isn’t there considerable opportunity for smarter transhumans, along with AI tools, to vastly improve the lives of many people by finding ways to mitigate problems we’ve inherited (disease, etc.) and created (pollution, conflict, etc.)?
Artificial intelligence (AI) is being incorporated into an increasing range of engineered systems. Potential benefits are so desirable, there is no doubt that humans will pursue AI with increasing determination and resources. Potential risks to humans range from economic and labor disruptions to extinction, making AI risk analysis and mitigation critical.
Specialized (narrow and shallow-to-deep) AI, such as Siri, OK Google, Watson, and vehicle-driving systems acquire pattern recognition accuracy by training on vast data sets containing the target patterns. Humans provide the operational goals (utility functions) and curate the items in the training data sets to include only information directly related to the goal. For example, a driving AI’s utility functions involve getting the vehicle to a destination while keeping the vehicle within various parameters (speed, staying within lane, complying with traffic signs and signals, avoiding collisions, etc.).
Artificial general intelligence (AGI or GAI) systems, by contrast, are capable of learning and performing the full range of intellectual work at or beyond human level. AGI systems can achieve learning goals without explicitly curated training data sets or detailed objectives. They can learn ‘in the wild’, so to speak. For example, an AGI with the goal of maximizing a game score requires only a visual interface to the game (so it can sense the game environment and the outcomes of its own actions) and an ability to interact with (play) the game. It figures out everything on its own.
Some people have raised alarms that AGIs, because their ability to learn is more generalized, are likely to suddenly surpass humans in most or all areas of intellectual achievement. By definition, once AGI minds surpass ours, we will not be able to understand much of their reasoning or actions. This situation is often called the technological singularity–a sort of knowledge horizon we’ll not be able to cross. The concerns arise from our uncertainty that superintelligent AIs will value us or our human objectives or–if they do value us–that they will be able to translate that into actions that do not degrade our survival or quality of existence.
• Demis Hassabis on Google Deep Mind and AGI (video, 14:05, best content starts a 3:40)
• Google Deep Mind (Alpha Go) AGI (video, 13:44)
• Extra: Nick Bostrom on Superintelligence and existential threats (video, 19:54) – part of the talk concerns biological paths to superintelligence
• Primary reading (long article): Superintelligence: Fears, Promises, and Potentials
• Deeper dive (for your further edification): Superintelligence; Paths, Dangers, and Strategies, by Nick Bostrom
Members may RSVP for this discussion at https://www.meetup.com/abq_brain_mind_consciousness_AI/events/234823660/. Based on participant requests, attendance is capped at 10 to promote more and deeper discussion. Those who want to attend but are not in the first 10 may elect to go on the waiting list. It is not unusual for someone to change a “Yes” RSVP to “No”, which will allow the next person on the waiting list to attend. If the topic attracts a large wait list, we may schedule additional discussion.
Members of this site who can’t attend the meeting are welcome to participate in the extended discussion by commenting on this announcement.