The last few months can simply be characterized as the “advent of AI” finally in the mainstream world and with that, a floodgate has opened that has brought with it a series of fascinating things. I have been interested in this area for some time and have commented on it on social media. I thought it might be best to write an article to cover various sightings in one place. I’m writing this article to look at this in an optimistic way, accepting the reality that these things are here to stay and focusing on how they can be used for the better rather than scaring me or them. ignore them all together.
There are some important words that you will encounter in this world. AI or what is called artificial intelligence is the ultimate goal of a program or set of programs that can not only behave like a human being, but rather process and think like a human being. For this, what we have at the moment is GPTs or pre-trained generative transformers and LLM (Major language models). Then there is Reinforcement learning from human feedback (RLHF) and a whole bunch of other keywords like quantification, tokenization, neural network, weights and biases in learning models, natural language processing. Now I don’t want to go too deep into these terms because each term can be expanded to fill an entire book or collections of books. I wanted to focus on how this shapes our new world. If you are new to this world but want to delve deeper into the technical side of the equation, I suggest you start from this article by Stephen Wolfram
What ChatGPT does…and why does it work?
Great language models and search engines
A notable shift is occurring in the way we seek information from search engines to interact with large language models like ChatGPT. Traditionally, the goal of search engines has been to be as succinct as possible, condensing your query to a few select keywords to avoid diluting search results. With ChatGPT, we find ourselves elaborating further, providing more context to refine potential responses. This comes from the fundamental difference in how we perceive these technologies.
Search engines are treated as tools – we direct them with precise instructions, while ChatGPT is seen as a system for understanding the language we converse with. The conversational style allows users to provide more details, which helps the AI generate more accurate responses. Additionally, search engines have been manipulated over time with SEO strategies, while AI systems have not yet been widely leveraged in the same way. Finally, while search engines present a variety of options, AI models give a direct answer. This is why we try to ensure that our contribution is as clear and complete as possible in order to achieve the most accurate result.
Reference:
Education and Generative AI
One of the key areas where big language models can revolutionize operations is in academia. Instead of sticking to conventional teaching methods, teachers could harness the potential of tools like ChatGPT to improve student learning. By encouraging students to explore topics using ChatGPT, teachers can transform traditional classroom dynamics. The first half of the course could involve students sharing the results of their AI-guided research, while the remaining time can be spent verifying the accuracy of these results, clarifying doubts and explaining any misconceptions.
What we need to understand is that although what I am suggesting is learning in isolation, one thing I have realized over time is that people like to learn in public but prefer to make mistakes or be aware of the mistakes they made in private. . Plus, it hurts a lot more when a human finds your mistakes than when a program finds your mistakes, you fix them and move on. Therefore, these LLMs provide an interesting and safe playground to make mistakes and learn from them.
It is crucial to understand that AI systems are not infallible: they do not have all the solutions and can sometimes have difficulty with certain queries. This is the perfect opportunity for teachers to intervene by asking thought-provoking questions to gauge a student’s understanding of a topic rather than just their ability to remember facts. Such an approach exploits the advantages of AI while recognizing its disadvantages. It’s a balancing act: as the saying goes, AI is a good tool but a bad master.
Reference:
Additionally, I would definitely recommend watching this video of Sal Khan talking about his adventures with combining AI and education.
Privacy and Generative AI
As I suggested above for students, it’s an interesting playground to ask questions and get answers without being judged. This would immediately get people interested in using these systems for a few use cases, including discussing personal issues or trying to fill in as a therapist. Please note that I am not suggesting the use of AI/LLM to ask anything and everything. you must exercise caution. Use the system, don’t give out the information you end up being used by the system. Interesting reading would be these links here, here, this and this.
Another interesting use case I’ve seen people explore is opening their thoughts or notes to AI and trying to extract insights from them. This seems like a good idea, but be aware that the data goes the other way, although OpenAI has ensured that data provided via the API will not be used to train their model and they have provided the ability to deactivate them via the interface. people need to exercise caution. What I keep saying: “Once data is made public, it should be considered public. » So decide before sharing your data with a public system.
OpenAI fixes account takeover vulnerabilities in ChatGPT
I’m not saying we’re not leveraging AI, all I’m saying is that maybe now isn’t the time to leverage a SaaS API to pass your data to someone other. There is a lot of work going on to get these systems running on individual machines, some of the efforts are listed here: and to list the main two I’ve seen, but there is a lot of activity going on in this area , so keep your eyes peeled. a better and safer solution would be available soon. is an interesting subReddit to watch for such innovations and developments.
Reference:
Rapid engineering and a shift in our thinking
I recently read this interesting article by Martin Fowler about his discussion with Xu Hao, an interesting observation that highlights the parallel between rapid engineering and the ability to clearly document thoughts. If we become skilled at designing effective prompts, it inherently improves our skill at clearly documenting our thoughts, and vice versa.
This observation highlights a potential benefit of engaging with large language models, namely the potential to improve the quality of documentation. This could potentially address a long-standing challenge within the IT industry: the need for better documentation. The connection between rapid engineering and clear documentation, while not explicitly stated in Fowler’s article, becomes evident upon careful consideration and carries significant implications for IT practices.
Reference :
LLMs and what they say about the world
Finally, I came to an interesting realization about the LLM and what this sudden boom also reveals about the inner workings of our world. Most important is the extent to which our world operates on repetition. The exams primarily test rote learning and the ability to memorize information. Additionally, intelligence is often equated with the ability to regurgitate information coherently. We test recall capabilities, which is why computers always seem smarter.
Reference:
Large language models like ChatGPT give an illusion of intelligence due to traits that we, as humans, typically associate with intelligence. These include eloquent sentences with minimal grammatical errors, words and phrases that seem to make sense, common but overlooked aspects of our environment, and the ability to quote eloquently without necessarily relying on them. stick to the original verbatim. Essentially, the emergence of intelligence in AI relies largely on our perception of intelligence itself.
Reference:
I think it would be safe to say that we all need to adjust our own understanding of what we call intelligence, take advantage of the barrage of tools and floodgates of technical innovation that these new technologies reveal to us all and can -be doubling what makes humans unique and “intelligent”. How we adapt to these changes and harness their potential will determine their effectiveness. As with all emerging technologies, the path forward is uncertain but full of potential.