For some years now, we have been permanently trying to attack our own business model in order not to miss any trend that could become dangerous for us in the future.
Especially in our field of numerical simulation, a lot is happening. The problem is that those who are not able to adapt to the conditions and requirements of the market will first wonder where the well-paid orders are, then sell on the price and eventually get eaten up.
I think we at Merkle & Partner play this game quite well. That's why we also keep an eye on what's happening in the field of AI and how we can use trends for ourselves in time.
Back in October 2018, Chadi and I were at a startup fair in Stuttgart and we talked to Sophia from Hong Kong, an AI who was a very convincing panelist.
Even though it seemed that there was something like intelligence at home, it was more like a trained monkey that plays its admittedly complex programme.
If you compare biological and artificial neurons today, nature is still far ahead. 100 billion with about 100 trillion synapses are compared to 10 - 1000, some even speak of up to 1 million artificial neurons.
While a neural network processes data serially, neurons in the brain fire in parallel.
Hardware wears out, the brain can repair itself to a limited extent. The brain is also more error-tolerant, whereas an AI fed with the wrong data produces pure data rubbish.
In terms of energy consumption, the human brain is also much more efficient at 20 W than an NVIDIA Titan RTX at 280 W.
So is it all just a matter of time before the computing power of modern computers overtakes that of humans?
I don't think so. There is a quality in life, let's call it spirit, that is more than pure biomechanics. If we just look at how many sensory impressions, memories, emotions, images and experiences we can remember in a human lifetime, it exceeds the storage capacity of even the best brain, especially since supposedly only 10% is used. I believe, contrary to current scientific doctrine, that this mind is superior to the material universe as we can study and measure it, and I am in good company with some renowned natural scientists.
But this is a philosophical question that pushes pure natural science to its limits.
But what AI can do is recognise patterns in contexts and apply algorithms. But that has nothing to do with understanding.
Here it is already far superior to humans.
In everything that can be captured in algorithms, that involves the evaluation of large amounts of data, everything that can be automated, AI will increasingly permeate our lives.
Being artistic and creative, dreaming up things that don't yet exist and making them reality and, above all, making other people laugh, is reserved for us humans. In contrast, Sophia's feigned empathy seems clumsy and empty.
So where is the AI sawing away at the computational engineer's chair?
In the foreseeable future, AI will certainly be able to create a nice computational network based on 3D data that produces good results. Performing work that can be automated, faster and cheaper than the most efficient engineer, is not an issue.
Even projects currently running in our company, in which the stresses and deformations are predicted using training files, already show a global hit rate of 98% with a few 100 samples (see Figure 2).
There are many tasks, even in computation, that have individual differences from computational model to computational model, but where simulation is too slow and too expensive.
This is where AI can play to its strengths and is already sawing away at chairs without those sitting on them noticing.
The calculation is simple: an individual component that costs less than a simulation for it, but that is often designed in a similar form, can be analysed and optimised in a fraction of the time with the right approaches.
Do you have such topics? Give us a call!
Your Stefan Merkle
PS: Socks with the corresponding scan of a foot with the I-Phone would fall under this heading.
PPS: No AI can replace a good calculation engineer in the foreseeable future. A monkey that works through a checklist can!