What are some impressive new AI projects?

If we look at real projects, the following is true:

  • Apple had hired an external company specially for their graphics chips, Imagination.

Unexpectedly they got the graphics chips back in the house. The reason is not far to look, the parallel computing power of both graphics cards and KI or AI chips is huge. Something so crucial, the graphical side and hardware acceleration for KI (AI) algorithms should remain within Apple’s walls. Apple has a history with power Pcs, after which they switched to Intel because they weren’t really a hardware company. Today they take ARM processor licenses (iPhone and iPad processor chips) and they develop their own chips, GPU and accelerators themselves. Hugely important because the effects on Intel and other generic processor manufacturers are going to be great.

  • Facebook is pure software, always been.
  • Well they now have their own hardware teams (ASIC and FPGA designers). Their own AI algorithms that dig through our data into better ways to convince us to buy things, that must remain secret. For their very specific algorithms, only tailor-made hardware can deliver the fastest implementation. A pure software company that develops hardware that shows what to expect in the future.

  • Google is a pure software company.
  • Their datacenters were based on common server processor chips. But to train their KI to beat the Go champion (a well-known game) and to get into Google photos (already wondered why photos offers free storage space, to train their KI of course) to recognize people and animals, they do not need general processors. They need very specific chips that can perform their algorithms the fastest and most efficient. Therefor they developed the TPU and eTPU, tensor processing units. The hubris of a software company fortunately made sure that their hardware is not really top, but that is a personal opinion.

  • Amazon is of many markets at home, especially automation of warehouses and delivering parcels.
  • But they also have their own chip now and through one or more hardware startups (Habana) More irons are in the fire. They were also the first to offer FaaS, FPGA as a Service, namely reconfigurable chips that can accelerate KI.

  • Microsoft is also not a strong name in the semiconductor sector and everyone knows what happened to Nokia.
  • Yet Microsoft is now top in the world in terms of “cloud computing”, even they have hardware teams, the future of MSFT is not Windows or Office but the cloud.

    Today, Asics and Fpgas are the future of the AI algorithms.

    AMD, Intel, NXP, Qualcomm,… will not continue to deliver to the big KI giants as these are going to develop more and more specific own chips.Their sales market decreases thereby. Plus The Giants are not very transparent and often appear to lie about their deeper motivations. Let them develop software and hardware without any supervision that will come to us within ten to twenty years of life.

    As for the first applications to be rolled out:

    1. Write content.

    OpenAI, an unclear collective of a gang of Klojos recently wanted to make their trained model to generate content not public because it was so good to write in fake news. No one ever went to their website to read their paper once. It stated that they had taken their dates from Reddit, about Brexit, Game of Thrones and Miley Cyrus. Just like any poll for voting behavior, this is also a classic for those who ever learned statistics. A sample stands or falls with the representative of the steel. Zever about Miley Cyrus is and remains zever, even if you release KI. Brexit is about pro and contra arguments that do not necessarily contradict each other but where one considers one more important than the other. And Game of Thrones is a fictional story, completely fabricated. An example of clutter in, clutter out. And no one who gives a kik, though. Their model trained on nonsense thus generated stories that were also nonsense and that was oh so dangerous. Just read the newspapers today, over the knickers of the princess from here and the famous singer from there. Very popular, gossip leaves, Boulevard Press, nothing new under the sun. Software proves to be able to generate such fabricated stories. Again a ten investigative journalists without work.

  • Hr.
  • We all know the spirit-killing companies where we are trapped all day in a noisy open space with all the colleagues. Where office politics takes you beyond anything. Where people are thrown outside because the profit was not bigger, but was just as big as the previous quarter. Well, those fine people who regulate all of this for us, the recruiters and personnel, who have a lot of benefit in analyzing data from employees. What kind of person is a hazard to the company? Which person remains faithful to the company even though he or she gets nothing back? What unscrupulous person can kick it to the new CEO. That kind of business. Completely coincidentally I read today in the Wall Street Journal the following (screenshot of WSJ dot com):

    Subversive elements will no longer be “manageable”.

    Most positions are searched for the largest mop with whom the entire office can be wiped.

    And as the WSJ puts it, there’s nothing you can do against it.

    KI is not good or bad, it’s statistic.

    It is the person who selects the data, chooses the algorithm and puts a desired outcome first.

    As always, it is man himself who is taking other people.

    Not in Disneyland, but in the rest of the world.

    Leave a Reply