Chat Generative Pretrained Transformer ‘ChatGPT’ v3.5 – a machine learning algorithm, trained on texts with billions of web-based content and interacts in a conversational way by OpenAI was released on 30th Nov. It is built on Microsoft Azure Cloud.
Within 5 days of its lunch, it reached the coveted 1M+ user benchmark.
It continues to be an internet sensation and could be how the year 2022 is remembered.
It has been used to write biographies, poetry, tweets, and even working software code.
Using its dialogue format, it comprehends the natural human language to respond in a human-like language.
It has been a fascinating preview of what is to come when we train machines – they could do things quicker and sometimes better than humans.
What Makes It Special, ‘Game Changing’?
ChatGPT gives seemingly informed, detailed, synthesised and creative responses.
The previously widely held opinion was that Artificial Intelligence/ Machine Learning ‘AI/ML’ would have minimal impact on knowledge workers with highly intellectual or creative jobs.
The expectation was digitalisation would mainly impact manual and repetitive tasks.
We had also thought this level of digitalisation, advanced computing power was a long way in the future.
ChatGPT has debunked these theories 🤯
It is also the first time advanced computational power is available in a very user-friendly, conversational way that is accessible to non-technical users.
So, Have the Machines Now Finally Arisen?! 🤖
Has the post-human era commenced? Does this mean we are now all redundant?
I think not quite yet.
There is still a need for humans to verify AI-generated output to ensure it is fit for human consumption.
Machines would probably take over the bulk of the initial research work.
And then need highly qualified humans to handle the verification of the content to ensure its fit for human consumption.
In other words, AI compliments the strengths of human intelligence.
However, this still leaves open questions on the future of work.
Questions like, these initial, research tasks were previously done by graduate, entry-level colleagues, would machines make these colleagues redundant?
Even more interesting – what new jobs would be created? How can we prepare for them?
Most of this is still unknown and evolving rapidly. What we do know is that change is coming and we need to get and stay ready by staying current with technology and being open to new opportunities.
But We Have Had Other Text-Based AI tools (e.g. Alexa, Siri) – How is ChatGPT Any Different?
1) ChatGPT can perform complex tasks that require creativity e.g writing jokes, speeches etc.
Check out this speech generated in seconds:
2) It has “memory”. It maintains chat history and can reference previous answers within the same chat.
E.g. I asked “Who is Beyonce?”, Then I followed up with “How many albums does she have?”
3) It questions or corrects incorrect assumptions in questions.
E.g., “When was the second time England won the world cup?” For comparison, I asked the same question on Google Assistant and incorrectly got 1966.
4) It admits its limitations and mistakes. It can refuse to answer questions. It has a semblance of “morals”.
1) It was trained primarily on web-based data available until the end of 2021. It has no information on events from 2022 and beyond. It is also not able to browse the internet to access current information.
2) The garbage in, garbage out (GIGO) principle still works. The user needs to know enough to ask the right questions.
3) It may mimic existing art but is yet to be able to create an original piece of art.
4) Although authoritatively sounding, some responses may be wrong.
ChatGPT is still in its beta (test) stages – its responses are usually a good starting point but need to be verified by humans.
5) For improved usability, it is trained to prioritise answers humans are most likely to ‘like’. This could lead to unintended bias for likes vs factual responses.
Five Possible Applications of ChatGPT
1) Great brainstorming tool – a more sophisticated search engine. It appears to be a very worthy rival to Google Search. It performs initial research and provides a good shell, starting point or framework for complex, creative work e.g. essays, movie scripts, music etc.
2) Code writing and prediction (The irony, AI automating software development!) – ChatGPT is already being used to write Twitter bots, WordPress Plugins, Text-based games etc. It can find & fix bugs in codes and even explain solutions.
3) Integrate with WhatsApp to get help from the tool – just like texting or phoning a friend but more instant and ever-present! An example of this service is “God in a box”.
4) Create initial strategy documents e.g., initial marketing plan, business plan, market research etc. Check out this sales page I generated on ChatGPT
5) Call centre automation – Not in its current form as it has limited data and can’t browse the internet. But eventually, could be used to provide human-like-sounding responses. Different from the traditional call tree.
One other impact I hope ChatGPT would have – is a reformation of the school syllabus.
Instead of pointless essay assignments that could be AI-generated (and potentially plagiarism-free too).
I hope this could be yet another push to change the syllabus making it more applicable to the digital era.
ChatGPT is based on open-source technology – meaning that enthusiasts globally can contribute to advancing the technology.
ChatGPT 4.0 is expected next year, 2023.
Any guesses on what the upgrades might be?
Have you used ChatGPT v3.5? What was your experience?