The source article is in Chinese. Click the picture to be directed.

We have summarized a very interesting article that highlights the most recent AI development as well as its future. Most of the information and comments are from this link which we translated. AI will become very powerful in every aspect, so the question for humans is how to apply it best in our world.

The next wave of AI: Studying the human brain 

This year’s Nobel Prize in Physics was won by Hopfield and Hinton. They used physics knowledge and tools to respectively invent the well-known artificial neural network, The Hopfield network and the Boltzmann machine, laying an important foundation for machine learning technology and promoting the development of AI.

Artificial neural networks are technologies that imitate the operation of the human brain. Hinton said that if we can understand how the brain adjusts the strength of connections between neurons, we can create amazing AI systems like GPT-4. AI will definitely be better than humans. It will be more intelligent, have a wealth of knowledge, but require far fewer neural connections than humans.

Hassabis, who jointly won the Nobel Prize in Chemistry with Baker and is the founder of Google DeepMind, believes that the next stage of research can use AI models to analyze the human brain to promote progress in the field of neuroscience. “I think this is a complete cycle. Neuroscience kind of inspired modern AI, and then AI will come back to help us understand what’s special about the brain.”

Can AI knowledge surpass that of humans?

When asked whether or not AI will be able to surpass the knowledge of humans, Hinton said that he believes we’ve already been surpassed. He added that though AI still has the problem of fabricating facts, also known as hallucinating, it doesn’t change the fact that it still has a greater range of knowledge than any human being.

When asked for his comment, Nobel Prize Winner in Medicine, Ruvkun, said that he believes human capabilities have been overestimated, and humans are just as liable to create false information like AI does. This is especially true with the rise of social media, which has only made the situation more serious. He concluded his thoughts by saying that it’s necessary for AI to reach the level of human beings, but it’s not a level that’s too difficult to reach.

Hassabis was also in agreement, and said that AI capabilities will advance rapidly in all aspects. Consequently, it is very important to think about how humans will design the AI systems and how they are applied in the real world.

What is a Hopfield Network?

A Hopfield network is a type of recurrent artificial neural network, often considered a “content-addressable memory,” which is designed to store patterns and can retrieve a complete pattern even when given a partial or noisy input, essentially acting like an associative memory system by minimizing an “energy function” to reach stable states; named after its inventor, John Hopfield.

Structure:

A single layer of neurons where each neuron is connected to every other neuron in the network, with symmetric connections (meaning the weight from neuron A to B is the same as the weight from B to A).

Function:

By adjusting the connection weights, the network can store patterns, and when presented with a partial or noisy version of a stored pattern, it will iteratively update its state to converge towards the closest stored pattern.

Energy Minimization:

The network operates by minimizing an “energy function,” where each state change reduces the energy, eventually leading to a stable state representing a stored pattern.

Applications:

Image restoration, pattern recognition, optimization problems, and understanding associative memory in the brain.

We hope this has been an interesting read. Now, what’s next? We look forward to seeing you in the next blog!

How the neural network works and defines the inputted numbers to produce an output is perfected by training the network with millions of sentence pairs. So for example, if you are trying to train a Neural Machine Translation engine for English to Spanish, you would need to feed the engine a great deal of data to help tweak and refine its framework and make it more accurate. Each sentence pair that is given to the engine slightly modifies the neural network while it uses an algorithm called back-propagation. Back-propagation consists of fine-tuning the error rate that comes from the previous iteration. By properly tuning, the error rates can be reduced and the accuracy can be improved.

So what is the advantage of using Neural Machine Translation? Some of the biggest limitations of other machine translation is that they have difficulty when it comes to more complex or nuanced phrases. However, with Neural Machine Translation, it becomes much more possible to translate these kinds of phrases since the number of parameters and rules that can be given provided are much greater and therefore it is more possible to generate translations that are much more natural sounding and closer in meaning.

Neural Machine Translation is without doubt the future of Machine Translation, but there are still reasons as to why it is not widely being adopted. The biggest reason is the sheer cost and time sink that the engines require to actually become useful. As mentioned earlier, millions of sentences need to be entered into the engine for it to start to really output translations that can be considered good quality, and that means using more manpower and time. Not every company is able or willing to invest that many resources to develop a Neural Machine Translation engine, and that is perfectly understandable when weighing the cost versus the reward.

Overall, Neural Machine Translation is a complicated system, but has the most potential out of all of the Machine Translation methods. It is very possible that Neural Machine Translation will be developed to the point where it can accurately be used in a variety of situations for which it is specifically trained, but it is unknown when exactly that might happen. There have been many advancements in its development, but due to its cost and time consumption it will be quite some time before it will be able to be commonly used.