With the current development of XAI and GenAI, there's the highest possibility of replicability of human intelligence or surpassing of human intelligence by AI. The fact that AI algorithms now are able to provide intuitive reasoning to back up their analysis and output makes these machines highly likely to surpass human intelligence in not too far future.

Possibly in the next year or two, the development of emotionally intelligent AIs could be developed with the ability to have consciousness as humans and generate synthetic feelings close to that of the real feelings of humans. Programming and training of AI is quite advanced now to achieve such feat. I'm futuristic on this thought, I'm espousing based on the current trend in AI development.

Touching on the ethical dimensions, I will like to raise the issue of non-acknowledgements functions of AI. The fact, that AIs are created and nursed on vast volumes of data from variety of sources without integrating into the system a way of the AI self-correcting and self-checking on thought patterns of ideas generated without acknowledging the prior state of the art makes AIs vulnerable for infringement and liability prone to works of arts and other knowledge sources that has to be duly acknowledged in the process of generating original ideas and crafts.

Notwithstanding, AI at it current state can be developed to surpass human intelligence and they'll be the possibility of creation of emotionally intelligent AIs. But developers and trainers could integrate within the training system of AI to self correct and self check to help not infringe on ideas, thoughts and crafts of others used for their training. With these ethical dimensions taken care of AI systems could become invincible and could make the works of human and intelligence obsolete.

Dear Tarik Houichime ,

AI's superior pattern recognition capabilities lie in its ability to process vast amounts of data, identify trends, and predict future outcomes based on past occurrences. While humans can do this to some extent, we are limited by the sheer volume of data that we can process and the biases we inherently hold.

https://www.linkedin.com/pulse/ai-better-than-humans-seeing-patterns-use-ld-graham-knowles/

Regards,

Shafagat

I think that there is a need to step back and consider several points about the advancements of neural networks and the gap between its output and human intelligence(as posted by George Benneh Mensah ) and human intuition/subconscious bias as the question poses. Let me address this before answering the question properly.

First, the question is about pattern recognition not about intelligence. There is a huge difference between the two. With respect to pattern recognition, a convolutional neural network nicely in complementing human pattern recognition in x-ray imaging[1-3]. That is not to say that they will replace doctors anytime soon. The underlying point here is that it does not replace the expert. In addition, the pattern recognition still needs some type of human filtering to make the NNs work for pattern recognition. There is still a need fro tweaking and preprocessing the data (even for LLMs there is a need to preprocess the data to remove HTML tags,avoid redundant data, etc.).

Second, to address the gap between pattern recognition and human intelligence, the neural network technology is based on generalizations and combinatorial closeness to derive results. Now I have not seen proof that this covers the universe of tasks that fall under human intelligence.

Now, the counterpoint to this last thought is that neural networks have been proven to be universal approximators. Anyone who reads the proofs will notice that even these proves comes with caveats making them existential and not universal in nature. A second counterargument is that, even though an algorithm can be a universal approximator it does not mean that it is feasible in reality. Take, for example, the Fourier series and the task of generating a wave with the outline of every city in the world and maintain its form as it evolves. Good luck in first finding the resolution on which to model this and make it universal and then finding the function that describes this evolving curve!!

Neural networks are wonderful, but as scientists we must see the phenomena not with excitement but with critical scientific evaluation. Let the excitement for corporations and their products. If we do this, then the scope will better fit reality and the ethical implications will be narrowed down. The societal impacts will lie in the applicability of the technology. In this aspect, initial expectations warp the reality of the impact.

References

[1] Medical Image Analysis using Convolutional Neural

Networks: A Review by Syed Muhammad Anwar et al.

[2] A Review of Convolutional Neural Networks for Inverse Problems in Imaging by Michael T. McCann et al.

[3] Convolutional neural networks: an overview

and application in radiology by R Yamashita, et al.