What sort of silicon brain do you need for artificial intelligence? The Raspberry Pi is one of the most exciting developments in hobbyist computing today. Across the world, people are using it to automate beer making, open up the world of robotics and revolutionise STEM education in a world overrun by film students. These are all laudable pursuits. Meanwhile, what is Microsoft doing with it? …
Wouldn't work around here. Our squirrels are used to heavy rain, they probably wouldn't even notice the sprinklers. Any countermeasures that would work reliably will probably put you on the most wanted list of our equivalent of the RSPCA or PETA.
Also: "a researcher saw squirrels stealing flower bulbs and seeds from his bird feeder" - why would you put flower bulbs in a bird feeder? They are way to heavy for the kind of birds you usually set up a feeder for. For watching them try to fly away with them and fail? Now that's not very nice at all.
Whatever happened to RAM based neural networks like WISARD?
When I first read about them in the late eighties they seemed to show great promise for visual perception and identification, as well as iterative circuit designs and other applications, but you never hear about them nowadays.
Are we in danger of overcomplicating things by using GPUs and algorithmic solutions, because we can, and forgetting what discrete, simple circuits could do?
We don't really need artificial intelligence, not when humans are still fairly common and cheap to hire.
What we actually need is things that are about as intelligent as a cockroach. That means an ability to find a way around obstacles, enough memory to get bored with going in the same circles all the time, and an ability to recognise simple dangers such as pitfalls and walls, etc.
Do this and do it cheaply, and higher-level functions such as navigation can be dropped on top from conventional programs. This sort of thing is already sort-of happening with robot vacuum cleaners, but needs to get better to be truly useful.
I'm pretty sure what you're describing is the exact definition of "artificial intelligence": any device that perceives its environment and takes actions that maximize its chance of success at some goal.
AI doesn't *necessarily* imply the instantiation of human intelligence in a machine/computer/otherwise 'dumb' object - although that is certainly one (common) interpretation. Of course, the definition de jour is (by its nature) constantly changing. By that measure, you're absolutely correct in asserting 'we're better off/more likely to develop cockroach-level intelligence than human-level intelligence'...for now.
Does "recognising a squirrel" represent "intelligence"?
More interesting would be to know if the device can distinguish a squirrel from a rat.
Even more interesting would be to know if the device improves its skills over time, say eventually learning to distinguish squirrels from rats without external help.
....all of which implies that the current marketing use of the word "intelligence" is just that...marketing.
Your Pi isn't going to be able to handle the differentiation between a squirrel, rat, mouse, vole, mole, etc.; I'm sure the "squirrel detection" scheme they're using is crude at best. Gonna need a higher resolution camera...and ideally the ability to detect changes in heat signatures and movement patterns. Sounds like a job for snickerdoodle. :)
I thought someone had finally figured how to build an artificial brain.
I struggle to see any AI in the article, or anything new. Just cost reduction due to process technology advancing.
I remember in 1980s someone told me that future more powerful CPUs would allow us to have AI.
I asked why wasn't there simply very slow, perhaps limited AI, maybe very expensive? "If you know how to do it, then a Z80 and external HDD could do it, slowly," I said.
In 1979, Cromemco released CROMIX, the first Unix-like operating system for microcomputers on a Z80. They updated it to 68000 during the 1980s. An early 1979 product was a digital camera for Altair S100 computer.
Very many years ago I was studying the neural anatomy of a flatworm. It has about 22 neurons. It is able to detect light and turn left or right to follow the light. It seems that is close to the level we are finally approaching.
What the new generations do not understand is that when it comes to AI speed means nothing. Some animals think slowly but still exhibit intelligence. A true AI might take a thousand years to have a new idea but it is still a new idea.
Your argument is flawless but for one detail : the marketing department will not be content with an AI that has only one idea every thousand years. It needs an "AI" that performs at lightspeed and wows everyone and their dog every second of the day.
At least during the TV commercials - once you get the actual product, you're on your own.
This post has been deleted by its author
No, the marketing dept need only to hide the fact that the AI can't perform at lightspeed and wow everyone and their dog every second of the day. Nor is even a single idea necessary - take another look at said marketing department.
Seriously, we won't know how to build AI until we understand the problem. And we won't know we have understood it properly until we have built an AI that really is. Comparing today's architectures is a bit like comparing flatworms with slime moulds (oh, leave the poor marketing department ALONE there, will you) and asking which is going to evolve into birds.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
