The video discusses the efficiency and energy consumption of artificial intelligence (AI) systems compared to the human brain. While the human brain is energetically efficient, consuming between 15 and 20 watts, AI requires significantly more energy, with training energy consumption comparable to the output of a nuclear power plant. This raises concerns about the sustainability of AI technologies as they continue to grow and evolve.

The video also explores the inefficiency of traditional silicon microchips used in AI systems, likening their energy waste to that of a candle's unwanted heat production. It introduces superconducting state technology and quantum computing, which offer potential solutions to these inefficiencies. Quantum computers, with the ability to control electron activity at a quantum level, present opportunities for vastly improved computational speed and energy efficiency.

Main takeaways from the video:

Please remember to turn on the CC button to view the subtitles.

Key Vocabularies and Common Phrases:

1. energetically [ˌɛnərˈdʒɛtɪkli] - (adverb) - In a manner that involves energy or forcefulness. - Synonyms: (vigorously, dynamically, powerfully)

But do you know how energetically efficient are artificial intelligence?

2. silicon microchip [ˈsɪlɪkən ˈmaɪkrəˌtʃɪp] - (noun) - A small piece of silicon with electronic circuits used in computers and other electronic devices. - Synonyms: (semiconductor, integrated circuit, chip)

Artificial intelligence is running on this silicon microchip that has a pretty new technology.

3. superconducting [ˌsuːpərkənˈdʌktɪŋ] - (adjective) - Relating to a material that can conduct electricity without resistance when sufficiently cold. - Synonyms: (supercurrent, superconductor, zero-resistance)

This is now known by quantum physicists as superconducting state.

4. operational speed [ˌɒpəˈreɪʃənəl spiːd] - (noun) - The rate at which a system or machine performs its functions. - Synonyms: (performance rate, processing speed, computing speed)

Promising operational speed thousands of times faster than the best modern computer.

5. quantum computer [ˈkwɒntəm kəmˈpjuːtə] - (noun) - A computer that makes use of quantum-mechanical phenomena such as superposition and entanglement to perform operations on data. - Synonyms: (quantum processor, quantum computing device, quantum system)

This is essentially at the base, very roughly speaking course, at the base of a quantum computer.

6. sustainability [səˌsteɪnəˈbɪlɪti] - (noun) - The ability to be maintained at a certain rate or level, especially without adversely affecting the environment. - Synonyms: (endurance, viability, eco-friendliness)

The sustainability of this technology may become a big issue to solve.

7. supply chain [səˈplaɪ tʃeɪn] - (noun) - The sequence of processes involved in the production and distribution of a commodity. - Synonyms: (distribution network, logistics network, supply network)

Going from logistic the supply chain solution or for the risk assessment for business.

8. energetic point of view [ˌɛnərˈdʒɛtɪk pɔɪnt əv vjuː] - (phrase) - Perspective focused on the use or efficiency of energy. - Synonyms: (energy perspective, efficiency view, energy consideration)

We can easily say that from the energetic point of view, silicon microchip is still at the air of a candle.

9. global energy consumption [ˈɡloʊbəl ˈɛnərdʒi kənˈsʌmpʃən] - (phrase) - The total energy used by all the people in the world. - Synonyms: (world energy use, international energy consumption, worldwide energy demand)

The big data server in which artificial intelligence is running are already consuming more than 1% of the global energy available.

10. quantum physicists [ˈkwɒntəm ˈfɪzɪsɪsts] - (noun) - Scientists who study quantum physics, the branch of physics relating to very small particles at the atomic and subatomic levels. - Synonyms: (quantum scientists, quantum researchers, microphysical scientists)

This is now known by quantum physicists as superconducting state

A superconducting waltz for tomorrow'squantumcomputing - Elia Strambini - TEDxForteDeiMarmi

Thank you very much for the nice introduction. The talk will be in English, so I hope you can follow. So, artificial intelligence is a very powerful tool that is nowadays in our hand, and we are quickly learning how to use it and take advantage of it. But do you know how energetically efficient are artificial intelligence? Respect to our biological one, the comparison is not simple, but something can be done. For example, we know that our biological brain consumes between 15 and 20 watts continuously. That amount of energy is equivalent to the energy consumed by an energy saving light bulb. So the association between light bulbs and EDEA is very well fitting also from an energetic point of view.

On the other hand, artificial intelligence is working on demand, and for each request, it typically takes between three to 300-watt hours, according to the complexity of the request, if it's a simple search or a complex drawing. But even more powerful is the energy that is required for training. Training that is pretty natural for our biological brain, can be very energy-consuming. In fact, we estimate that it takes around a billion watt-hours. That is equivalent to the energy produced by 1 hour by a big nuclear-powered plant. And so the continuous amount of requests that we daily do to artificial intelligence and the periodic training that they require to be updated makes the energy consumption of nowadays artificial intelligence equivalent to energy consumption of countries like Ireland.

And we are just at the beginning. The new generation of artificial intelligence will come soon. They will be more powerful, but will also consume more energy. And they use artificial intelligence continuously, continuously growing. And so, considering that nowadays, the big data server in which artificial intelligence is running are already consuming more than 1% of the global energy available, the sustainability of this technology may become a big issue to solve.

But why artificial intelligence is so inefficient with respect to our biological brain? We know that a biological brain, if you think, is a beautiful computing machine coming from millions of years of evolution, and where energy saving was one of the requirements for natural selection, on the other hand, artificial intelligence is running on this silicon microchip that has a pretty new technology that was developed in the 20th century and based on performances rather than on average, energy saving. So if we associate the energy-saving light bulb to our brain, we can easily say that from the energetic point of view, silicon microchip is still at the era of a candle. And like the candle, most of the energy that you put in your microchip is producing unwanted heat.

And just very few of it is really used for the computation. And that's something that you experience every day. How many of you have never felt your mobile phone or your personal computer burning after a deep use? And do you have the same feeling with your brain? Have you never had a high fever after a deep thinking? Maybe a headache, but not a burning brain? And that's essentially the point of the inefficiency of artificial intelligence. And this problem is so acute and timely that nowadays, the big corporation of high tech are building their data server in poorer region or under the sea, just to save money on cooling costs. And this is a real example of the funkoil that are using to cool down this data server. And while this solution may find maybe the economic needs of this big company, this for sure will not solve the sustainability problem of this technology.

So a new deep change in the foundation of how we build up the future data server is needed. Can we design a new computer that is outperforming our modern computer and saving a lot of energy? So it's consuming much less energy. To address this question, it would be nice if we could enter in a microchip and have a look at the atomic scale. If we could do that, this would be the scenario in front of us that's very familiar with this location. I think it's a big party with million or million of electrons dancing. And in this metaphor, the high disco music that the electron follow is just the temperature of your microchip. Does it look very efficient? If you now ask a calculation to your microchip, then one of the electrons has to play as a waiter and has to take your request to the bar. To do that, it has to crawl across all the busy, dense floor. This process is very slow, as you can imagine, and inefficient.

It will be shaken by the electron. Additionally, increasing the temperature of your microchip. Once arrived at the bar, it will process your request. In this example, just a simple cocktail that needs to be sent back to you. Sent back to you again by crossing the dance floor. And if you're lucky, you will receive the cocktail. Otherwise, you have to make the request again. So how can we improve this process? It would be beautiful if now a mobile phone, we could have a stop party. If we remove all the electrons from the dance floor, we can improve the efficiency. And how do we stop a party? People that are familiar with disco, they know for sure the best way to stop a party is to stop the music. No music, no party. And so a way to stop the music in this example is to lower just the temperature of the processor.

And indeed, if we lower the temperature of silicon microchip, we can improve their performances. But then another problem came out. When the temperature is too low, no electron will be there, including the waiter. And the computer stopped working. But luckily, physicists discovered more than 100 years ago that that's not true for all the material. There are a lot of material very common, like aluminum, vanadium, niobium and many more on the study, which when you stop the temperature, the music, something special happens. And in fact, when you lower the music, you hear that there is a sound in the background. And then you discover in this material that in one of the corner of the disco, there was a big orchestra playing a beautiful valse music, the electron, by following this waltz music, they are excited and they start to couple together and they start to change the way they dance in a beautiful, vast way, so coherently in a very ordered circle. This is now known by quantum physicists as superconducting state.

And now, if you make a request to your computer in the superconducting state, and the weather is smart enough, you can simply cross the dance floor by following the mainstream of the dancer and then back to you with your cocktail by following the backstream. And this is already known, and it has been proposed for the future supercomputer, promising operational speed thousands of times faster than the best modern computer and with an energy consumption that is even better than our biological brain, simply because the waiter will not be shaken by the electron, but can pass through it.

But the surprises are not finished, because the same quantum physicists discover a few hereafter also how to control the orchestra. And if you control the music, you can control all the dancers at once, and then you don't dine the service anymore, but you can use all the dancers coherently for your service with an efficiency that is even much, much more than the classical computer. And this is essentially at the base, very roughly speaking course, at the base of a quantum computer.

That's a technologically holy ground in which big countries are really investing a lot. US, Europe, China, and as well as big companies like Google, Microsoft and IBM, in this gold rush to the realization of the first quantum computer. And the first announcement already came with, for example, Google recently showcasing their Sycamore processor. This small 1 cm by 1 cm, comprising just 53 quantum bits, are already able to solve a complex problem in just 200 seconds. The same problem run on a classical computer would take 10,000 here. If we now we will be able to increase the size of this computer, we'll be able to release the full power of this technology, and we have benefit for business and society.

For example, with a quantum computer, you can simulate complex chemical reaction and have a near instant design of drugs. A process that nowadays takes years of lab activity for the selection of all the molecules. On the other hand, you could, for example, optimize the process of the fertilizer production. That is again, a problem nowadays is producing a lot of pollution. And more and more complex problems have been proposed to be solved, going from logistic the supply chain solution or for the risk assessment for business. So we do believe that we are at the dawn of a new era of computing in which we will have unprecedented computing capability while maintaining the integrity of our planet.

The future is exciting, but then when we reach such high level of computing, a question may come out. What are we going to do with our biological brain? Are we going to replace with an efficient solution supercomputer? I don't think so. Even with the best quantum computer available. In fact, all these machines, they have big advantages. That is also the strongest limitation compared to the human brain. They can sleep, so they can work 24 hours a day, but they can dream like human. And it's thanks to the dream of beautiful mind that we were able to achieve the advancement that we are in which we are now. So please remember to feed and take care of your biological brain.

Artificial Intelligence, Technology, Science, Quantum Computing, Energy Efficiency, Sustainability, Tedx Talks