I was reading an interesting book by the Physicist Michio Kaku called Physics of the Impossible. It has a thought provoking chapter in it on Artificial Intelligence and Robots.

In it he discusses the difference between syntax (grammar,structure) which is a logical rules driven exercise and semantics (meaning). He talks about how various influential scientists and futurists such as Roger Penfold believe that true Ai is impossible.

He also discusses the whole area of consciousness and its applicability to Ai. Interestingly enough he refers to himself as a constructionalist ie instead of debating if a true Ai is possible just get on and try to build one. I like that kind of thinking.

We are definitely not trying to build a true Ai at Virsona .. only one that appears to have intelligence. Kaku uses the example of a thermostat. It knows when the room is cold and can turn on the heater. Is that intelligence?..could the thermostat said to be an Ai? No clearly not; however it does imply that there is a spectrum on Ai ranging from the thermostat up to a true thinking machine.

As we continue to improve both the syntax handling capabilities and the semantic capabilities of our technology we will move closer to emulating an Ai,further up this spectrum if you will. Right now however it's still just logic, even if its really smart logic.

The leap to a true Ai is something we will have to put in next years development plan I think... or maybe the year after that.

Silicon Roses....

I was chatting with Einstein today at our chat site for him and he has this great line about "the only reason for time is so that everything doesn't happen at once".

Working in a tech startup everything does actually happen at once which is probably why there are never enough hours in the day.

As we continue to develop new capabilities into the Virsona engine we have been focusing on system resource allocation, processing and effective ways to enhance the conversation. One of the interesting subjects that came up is a standard issue that chess computers faced in the early days. A brute force approach, I can do more calculations than you in a faster time, was the initial way to be win.

While it is true that more calculations, faster can get you to a destination more quickly it doesn't necessarily mean that the answer is significantly better than one that requires less processing but takes more time to decide on the right paths to take.

Is there something inherent in the amount of time that we use to process thought that processors cannot emulate? The fact that we think in a specific way that is constrained by the chemical reactions and pathways that drive thought is in fact the key to our creativity and therefore throwing power and speed at the problem will never allow us to emulate it.

Perhaps the answer is to make our Ai engines stop and smell the silicon roses?

It's a butterfly....?

I have spent a lot of time thinking about personality recently. My head is swimming with everything from the Five Factor Model to Enneagrams to Social Cognitive Theories.

Turns out that personality is not an easy thing to quantify or even qualify exactly what it means. Psychologists seem to abound with differing theories of the why and approaches of the how.

As we look at adding personality to our Ai technology we have had to pick thru much of the work and look at what is not just useful but also practical from an implementation perspective. Personality models that use language as a basis make a lot of sense as that is somewhat more easily controllable..ie using a certain word over another is an indication of personality type and while we all do it subconsciously people 'assign' personality by listening to those words we use.

In the movies and books Ais always seem to have a strong fixed personality type ...usually quite sarcastic it seems. From Marvin the Paranoid Android in Hitchhikers to Robbie the Robot or even to my good friend Hal. However the reality is that as we start to see more 'personality' come into our hardware and software constructs these will not be fixed personalities but rather will blend to match the needs of the user.

As least they will if I can just figure out what that ink blot actually is...