A personal observation of the advent of AI (Artificial Intelligence) since my previous write-up in Feb: https://medium.com/@checkwoei/the-advent-of-ai-chatgpt-and-others-exploring-the-next-frontier-1bce9144aedf ChatGPT was only introduced to the world less than 5 months ago and the pace has accelerated exponentially in this short period. Many new inventive uses are being discovered and completed AI products are appearing on social media posts everywhere.
The genesis of an exponential rise in AI interest in recent times can be traced back to DeepMind’s AlphaGo defeat of the best Go player in the world in 2016. I highly recommend everyone to watch this documentary on youtube: https://youtu.be/WXuK6gekU1Y . It made everyone sit up and realize the potential of deep-reinforced learning. By 2017, China had declared that big data will be its new “oil” resource. It also kickstarted programs nationally to turbocharge its AI initiative since then.
While there is always the fear of jobs being lost to AI as machines take over, my takeaway is that we must learn to embrace and utilize AI in our daily lives. The alternative of ignoring it will be that we will be disadvantaged in our careers and be left behind. AI will improve our work effectiveness and free us to focus on higher value-added opportunities. Newer job opportunities will be created. While old jobs will become obsolete, many more new ones will appear. https://www.bnnbloomberg.ca/335-000-pay-for-ai-whisperer-jobs-appears-in-red-hot-market-1.1901850
At this point, I just like to share a bit of my learning journey and how I stumbled into AI. With the Skillsfuture government subsidies to help mid-career people pivot to growth areas like Tech, I started to embark on a new learning path in 2018. An old dog needed to learn new tricks to stay relevant. I wanted to reinvent myself while trying to pivot to a consultancy career at the same time. I was an ex-dark side banker looking to redeem myself with positive Jedi energies. Since then, I had completed various tech courses to educate myself. This includes a Diploma in Business Analytics, IBM’s SGUnited AI course, NUS’s Fintech program and various short courses that align with my goals.
The more I got into data analytics, big data, cloud computing and machine/deep/reinforced learning, the more convinced I am that this is the beginning of a brand new world where exponential growth is possible. The stars have finally lined up as computing power is now able to absorb and make sense of the huge amount of big data around us.
One area of interest that has progressed so rapidly within the last 18 months is the development and increased sophistication of deep fake image and video production. I remember doing a team project in 2021 for our IBM AI module on this topic. We were showcasing the fake Tom Cruise videos as an example of what was possible then. We used GAN (Generative Adversarial Network) and simple Python programming tied to public pre-trained datasets of less than 10k images to help spot and identify fake facial photos as a new account application security measure. Boy, they sure look crude and amateurish now compared to what one can do nowadays with a simple prompt and a click of a button.
Just look at the most popular generative AI applications like MidJourney and Stable Diffusion which can easily produce any realistic photo-like quality from anything you want via a few sentences. You have probably seen by now the video of the Pope wearing high-fashion Balenciaga outfits that fooled a lot of people. The originator did not need to draw well. He simply had to type words to prompt the generative AI to create realistic images.
With the combination of ChatGPT and these apps, the possibilities are limitless. One can prompt ChatGPT to fine-tune a detailed description to apply to the image-creating software to invent a photo-realistic art piece. We can also easily animate any person’s image to create a video of him/her speaking from any script we want.
Imagine what it will do to the advertising industry! There is no need to hire real people for new advertisements anymore. One can create personas who do not exist in real life to sell products without any fear of copyright violations. There are serious concerns that we will soon be seeing a bombardment of deep fake videos of well-known people hawking consumer products or spewing fake news.
There are also several new innovative ideas on how one can monetize all this new tech. An interesting suggestion was to develop online courses using ChatGPT to come up with full training modules and then create animated avatar lecturer videos to make the training interactive. The turnaround time would be hours instead of months thanks to the AI applications available now.
Another area of huge interest will be the medical field. AI data analytics is good at image detection and will be able to spot trends and patterns which the human eye is unable to. It was also used by banks to help spot fraud signals in credit cards where big data was available. For example, one can train the model on 10 million normal X-Ray images and separately train it on another 1 million images of positive cancer X-rays. Then when you show them new random X-rays, the trained neural network will be able to detect cancer with a high level of probability or indicate that it might happen in the future. https://www.nytimes.com/2023/03/05/technology/artificial-intelligence-breast-cancer-detection.html
To show how easy it is for a non-IT layman like me to do so, I used stable diffusion https://stablediffusionweb.com/#demo to create the following image below by typing a simple sentence: “Merge Joe Biden and Putin into one person, talking in Congress”. I can then prompt ChatGPT to do the following: “Imagine you are the presidents of the USA and Russia. Write a 300 words political speech on the following topic: To offer my sincere apologies to Ukraine and China for what we have done.”
I then use another freely available AI application to combine the face photo with the ChatGPT script to create a talking video. https://www.myheritage.com/deepstories/ The finished product will be of the imaginary person reading the whole speech like a normal human video with various facial expressions. The end-to-end process can be completed within 15 minutes. https://youtu.be/ZBSQZm7WrUs Fakes will get harder to detect as these AI tools improve exponentially in the near future.
The advent of AI is here and the possibilities are endless even as newer versions with bigger trained datasets are being rolled out within months. Let the games begin!
Leave a Reply