When will computers become smarter than humans?
If Moore’s law keeps continuing, there will be a point in time where computer processing power exceeds the processing power of the human brain. Faster computers could have a huge impact on everyday life and the tasks that we perform. To give a deeper understanding of how fast the brain we will use the estimation of the processing power of the human brain that has been made by Dharmendra Modha, IBM Fellow and IBM Chief Scientist for Brain-inspired Computing. He estimated that the brain has 38 pentaflops of processing power, which is a thousand trillion or 38,000,000,000,000,000 in numbers. Flops stand for floating point operations per second and is an indicator for the processing power of the CPU (Central Processor Unit). Some estimated examples to put the human brain into perspective:
Iphone 6 has about 6,250,000,000 flops.
Samsung S6 has about 33,000,000,000, flops.
Nintendo Wii U has about 333,000,000,000, flops.
Playstation 4 has about 1,833,000,000,000, flops.
Tianhe-2 upercomputer has about 33,860,000,000,000,000 flops.
As you can see the world’s fastest super computer’s computing power is getting close to equal the human brain’s processing power. But when will the commercially available processors surpass the processing power of the human brain? The fastest commercially available are the Core i7 5960X and 5930K, which have about 354 gigaflops (354,000,000,000). According to Moore’s Law (with the help of multithreading and service-orientated architecture) it would take another 32 years before processors faster than the human brain would be commercially available.
Processing power is one thing, modeling software to behave and think that surpasses human knowledge and rationale is another. Artificial intelligence is already being developed, but is nowhere near human intelligence. The combination of super fast processors and software that can improve other software could lead to exponential technology development. This could bring many benefits such as human augmentation, robots that will do a lot of human tasks and increased efficiency in everything that is computerized. We also have be cautious when the technology develops at an exponential rate, as artificial intelligence could “outsmart” human beings. By looking at the current trends in the development of processing speed we could definitely say that there are some exciting technological developments/revolutions to come.
References
Forbes.com, (2015). Forbes Welcome. [online] Available at: http://www.forbes.com/sites/alexknapp/2014/06/23/chinas-tianhe-2-remains-the-worlds-fastest-supercomputer/
Pages.experts-exchange.com, (2015). Processing Power Compared. [online] Available at: http://pages.experts-exchange.com/processing-power-compared
Puget Systems, (2015). Linpack performance Haswell E (Core i7 5960X and 5930K). [online] Available at: https://www.pugetsystems.com/labs/articles/Linpack-performance-Haswell-E-Core-i7-5960X-and-5930K-594/
Researcher.watson.ibm.com, (2015). Dharmendra S. Modha – IBM. [online] Available at: http://researcher.watson.ibm.com/researcher/view.php?person=us-dmodha
started, I. (2015). Intel processors: what you need to know to get started. [online] TechRadar. Available at: http://www.techradar.com/news/computing-components/processors/intel-processors-everything-you-need-to-know-1282987/3
11 responses to “When will computers become smarter than humans?”
Leave a comment
Recent Comments
| From Farmville to th… on The success of Estonia’s… | |
 | Tycho van der Scheer on How will traditional grocers c… |
 | 374482rm on Bitcoin The Disruptor |
 | VH on Makeup Genius: How L’Oréal is… |
 | Josephine on 3D printing: a new standard of… |
Information Strategy 
I think it is a very interesting concept you have captured here. I have found a related article which looks at this event in a totally different way. They argue that computers will never be able to exceed human intelligence, because of the way humans blend in emotion and reasoning. Computers run solely on algorithms which can solve very complex issues in just a second, but they will never produce a real breakthrough because they act predictable. This predictableness is an obvious result of the algorithms of course, but it is interesting to see that they do mention that some disruptive innovations have emerged out of the algorithmic brain. Stuart Kaufmann, a complexity theorist, explained this as the “partial lawlessness”, which is a little gap that allows creativity to come out of regulatory. (www.huffingtonpost.com)
My personal opinion on the subject more or less corresponds with the article I have mentioned above. Artificial intelligence is becoming more, and more influential. But I do not expect it to be able to fully replace the human brain. Of course, simple tasks can be performed by computers and more complex tasks can also benefit from these systems. But I do not believe the human brain can be replaced.
Author: Jessica Doper (358278jd)
Source: http://www.huffingtonpost.com/nick-seneca-jankel/ai-vs-human-intelligence-_b_6741814.html, accessed on 10/10/2015
Indeed an interesting you have covered! Your comment on how artificial intelligence could possibly “outsmart” got me curious. Can smart computers/robots truly threaten humanity in a I-Robot type of scenario?
As of right now there are already tasks that computers can do better because of artificial intelligence. Like a game of chess, worlds best chess player cannot beat the best chess AI. Also search engines, aircraft autopilots, fraud detectors and stock trading agents are all examples of smart AI’s. However there are certain things like common sense reasoning and thinking out of the box, which are impossible for any AI. That’s why AI’s as of right now have no way of threatening humanity by themselves even if their processing power would surpass the 38 pentaflop point.
Even though a doomsday scenario of hateful robots is not likely, smart AI’s still pose a threat in a different way. Many jobs might be taken over by smart computers, studies have shown that nearly 50% of the American workforce are threatened to be replaced by their computerised counterpart.
http://www.wsj.com/articles/does-artificial-intelligence-pose-a-threat-1431109025
To link myself to the previous comment, I do agree that computers with more processing power than humans will be created in the future, nevertheless they will hardly be able to outsmart humans.
On a more practical side this means that many jobs are truly at stake, especially those that requires the perform repetitive and mechanical tasks. I suggest here a quite interesting tool developed by BBC that enables the uses to assess the likelihood of his job to be replaced by a robot or a software (http://www.bbc.com/news/technology-34066941).
Substantially, this tool evaluate how much a job has those characteristics that will enable an advanced machine to perform it. Interestingly, those jobs that have the lowest probably of being replaced are those with the highest degree of creativity involved, those that require to work with other people and those that involve innovation and thinking.
Nice post, Pim!
I like that you reduced brain and computer processing powers to a common denominator of floating point operations per second, which allowed comparing the two with each other. Although I’m still confused on what exactly these “flops” mean. According to the Keil, “a floating-point operation is any mathematical operation (such as +, -, *, /) or assignment that involves floating-point numbers (as opposed to binary integer operations)”. But how exactly can human outperform the best existing computer? I believe that humans score high on solving complex tasks in terms of content, while computers score high on solving complex tasks it terms of the number of operations being executed. If you ask the smartest person in the world to make 1,000,000 simple calculations, such as 2+2, he will have to spend several days on doing that, while the most powerful computer (in terms of processing power) will handle the task in 1-2 seconds. Could you please explain in more detail what exactly a “flop” means when we are talking about brain processing power?
You’ve also mentioned Artificial Intelligence in your post, although didn’t elaborate much on it even though processing power and AI are highly interrelated. AI is indeed not standing near human intelligence, but it’s just a matter of time. Once more sophisticated AI programs will be developed, which will allow AI to actually learn and evolve, the processing capacity of computers available at times will be the ultimate constraints of AI’s development. Actually, it is interesting to speculate a bit and ask the following question: once AI will be able to learn and evolve, will it be able to find a solution on how to increase the processing capacity of computers? Moore’s Law holds for human’s contribution towards processing power improvement, while AI may be able to find new optimal solutions much quicker. Going back to the point, I would like to suggest to elaborate more on how raising processing capacity of computers will contribute towards AI’s development.
As a final suggestion to your post, I would recommend providing some real world examples of how raising processing power of PCs will affect certain businesses and industries. I am aware that high frequency trading in finance industry is a raising trend in nowadays. Computers are able to perform 10,000,000 trading operations in just fractions of seconds. Do you think Moore’s Law will change the way certain businesses operate?
References
http://www.keil.com/support/docs/1476.htm
http://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html
http://www.marketwatch.com/story/heres-the-advantage-high-frequency-trading-firms-have-over-everyone-else-2015-08-13
I find the article very interesting. I do think that computers will surpass the human brain one day in terms of calculation power. However, computers wont be comparable to human brains on the emotional level. It is true that emotional intelligence is possible in some way for computers: For instance, some machines are able to analyze the emotional state of humans by analyzing the pulse with a heart rate monitor and simultaneously scanning the face of people. However, there will be always a calculating logic behind these processes. Machines will never be able to feel; they will not be able to experience true emotions. Anyway, very intersting post!
There is something about “emotion” that I cannot agree with.
Human also have a calculating logic behind their process to analyze other people’s emotion. Smiley face means happy or tearing eyes means sad. Our perceptions of emotions are based on our experience of the external environment. The major difference is that human’s perceptions are not always that accurate, or sometime random, given the same external stimulation. So maybe someday in the future, we can build a machine which can mimic human’s emotional expression by implementing a less accurate perception model.
My question is this as following: If a robot can perfectly mimic human’s emotional behavior using a calculating logic, (a robot which can pass the Turing test) what is the difference between his calculated emotion and the so-called “true emotion”? Or does it even matter?
A well found subject related to nowadays technology and very interesting as well. However, you assume that Moore’s Law will continue, I did some research and came to the conclusion that we might reach a roof, to where Moore’s Law will not longer apply. This year fifty years ago Moore came up with a principle that became elevated to a law: Every two years the number of transistors on a computer chip will double.
Chips these days are still made from silicon and the fact that silicon scaling has less than a decade left to live, worries researchers. Faced with this deadline, chip manufacturers are investing billions to study and develop new computing technologies. If we want Moore’s Law to continue, researchers say we need a solution for this problem and not long from now.
Related to your subject; when will computers be smarter than the brain. Taking this into account it might take some extra time and even more importantly at what cost?
http://www.scientificamerican.com/article/moores-law-computing-after-moores-law/
The Guardian, 2014. ‘The question: Will Moore’s law fail in the next 20 years?’, http://www.theguardian.com/technology/2014/nov/09/the-question-will-moores-law-fail-in-the-next-20-years, visited at October 17, 2015
Interesting article. I do believe that computers will outsmart us in some point of time. It think it will be only a matter of time. Right now Google and NASA are working together on Quantum Computing. The computers that we know work with bits, which can only be in one state, either 0 (off) or 1(on). For a quantum bit is possible to be in two states at simultaneously. The ‘qubit’ can be on and off at the same time. This drastically increases the speed of the computer.
This ‘Quantum computing’ can be used by google to increase its machine learning. Which is a branch of AI that is crucial for the success of Google. Google wants to combine the extreme speed of quantum computing with its traditional data. Combined the computer can build more accurate models for every business of Google. For example, a Google user is looking for a new phone. The machine can now use all of the data, like previous searches, and find the perfect phone for us. It will develop to a certain point where the search engine knows us better than we do know ourselves.
Khomani, N. (2014), 2029: the year when robots will have the power to outsmart their makers, http://www.theguardian.com/technology/2014/feb/22/computers-cleverer-than-humans-15-years
Metz, C. (2015, Google’s Quantum Computer Just Got a Big Upgrade, http://www.wired.com/2015/09/googles-quantum-computer-just-got-a-big-upgrade-1000-qubits/
Tverdohleb, V. (2015), NASA and Google’s Quantum Artificial Intelligence Laboratory, http://www.itechpost.com/articles/15467/20150730/nasa-and-googles-quantum-artificial-intelligence-lab.htm
Extremely interesting subject.
I recently read a book by Nick Bostrom, one of the leading minds in the subject of Artificial Intelligence, in which three potential paths to achieving human-level AI are described:
1. The AI path
2. The Whole Brain Emulation path
3. The Neuromorphic path
The path that caught most of my attention was the whole brain emulation path. This path would involve creating a 1to1 model of the human brain by mimicking the low-level functions of the brain to such a detail that it would be able to produce the “phenomenological effects of a mind”. Emulation would solve many issues, including bypassing the need for extreme processing capabilities as described in your post.
In order to achieve such an emulation a several components would be needed. These include a complete 3D scan of the brain at a high resolution, a brain database including low level objects of the brain, and a functional brain emulation. These components would be then combined with the goal of eventually obtaining emerging properties of an individual brain including personality and social functions.
Bostrom even wrote a paper detailing the process to get here (check the link bellow).
It makes me both extremely curious and scared at the same time to know that such technologies would be possible (and that there is a roadmap to get there 😛 ).
Definitely recommend the read (its extremely dense, and I can’t remember many of the details, but definitely mind blowing):
Book – Superintelligence – Nick Bostrom
Whole brain emulation:
Click to access brain-emulation-roadmap-report.pdf
Review by the FT
http://www.ft.com/intl/cms/s/2/021d3484-fd1d-11e3-8ca9-00144feab7de.html#axzz3ovU44k5R
Great article! I have to admit that the idea of combining very fast processing capabilities with artificial intelligence is both exciting and frightening. Yes, it would help us tremendously in our daily lives, having some sort of personal aid that could take care of basically anything that doesn’t require a physical action (unless we combine it with robots of course). However, such ability, from what remains a machine, must be controlled in order to prevent any type of issue (I am sure you can think of some…).
Luckily, as you mentioned in the article, artificial intelligence is not developed enough to offer anything worth getting super excited about, which gives scientist plenty of time to evaluate the pros and cons of such technology.
Finally, I would like to recommend a great TV show in relation to AI: Person of Interest. I believe it really demonstrates the issue revolving around artificial intelligence.
Intriguing concept you’ve covered. However the number amount of connections flops, connections or synapse if we talk about human doesn’t directly has to do something with outsmarting the other.
To outsmart the other party a certain intelligence has to be accomplished. If you go deeper into the concept of intelligence Alex Wissner found in his research(1) that it can stated that intelligence has to with creating maximize future freedom of to say to keep options. The more options there are, the better. So it the moment the machine can learn how to improve myself, this will increase the number of options it has. The moment it will give the machine an advantage to outsmart the human to provide better options of future freedom it will start to develop in that direction. However that kind of learning is a trial-and-error form of learning. So the moment the machines will outsmart us, we’ll be aware of it and hopefully by that time we’ve also developed ourselves. Maybe we can use the technology to improve also our capacity. This could be in the form of external storage of information, memories and emotions, by improving our senses or by be better in math. So it will be in interesting concept if the machine is ever going to outsmart of or that we eventually can keep up with the machine.
329761
(1) https://www.ted.com/talks/alex_wissner_gross_a_new_equation_for_intelligence#t-346454