Everybody has a different opinion on ChatGPT, and AI. It is a frontier for engineers and entrepreneurs, a place to create new products, services and solutions. Journalists and social scientists are concerned, with Ezra Klein , a prominent NYT author, calling it an “information war machine.”
Let me be clear, I see tremendous possibilities. We cannot predict the future impact of new technologies yet, as we can’t fully anticipate them. While there will be failures and problems, the end result is always “hooray!”
What is ChatGPT?
Table of Contents
This technology (and many others like them) is commonly known as a “language engine” and uses statistics, reinforcement learning and supervised learning to index sentences, words, and phrases. Although it doesn’t have any real intelligence (it doesn’t know what a word “means”, but it does know how it is used), this technology can answer questions, create articles, summarize information and much more.
ChatGPT engines are “trained” (programmed, reinforced), to mimic writing styles, avoid certain conversations, and learn from you questions. The more advanced models can answer more questions and store the information for future reference.
This isn’t a new concept. We have had chatbots for over a decade. It answered questions such as “what are the best practices to recruit” or “how do I build a corporate training program” with good results. Although the answers were a bit basic and a little inaccurate, they will improve with training.
It also has many other capabilities. It can answer historical questions (Satya Nadella is the president of the United States in 1956), write code (Satya believes that 80% of code will automatically be generated), and write news articles and information summaries.
One vendor I spoke with last week uses a derivative GPT-3 to automatically create quizzes from courses and serve a “virtual teaching assistant.”
(PS: In some ways, the chatbot may be a commodity. At most, 20 startups have highly-funded AI teams creating competing or derivative products.
What Can ChatGPT Be Used and Similar Technologies?
Let me tell you why I think this market will be so huge before I go into it. The corpus (database of information they index) is what “trains and educates” these systems. GPT-3 has been trained using the internet and a number of highly validated data sets so that it can answer almost any question. It’s kinda stupid, in a sense, since “the internet” is full of marketing, self-promotion and news. I believe we all have enough trouble figuring out what’s real. Try searching for information about your latest illness to see what you discover.
Google’s competitor to GPT-3 (which is rumored be Sparrow), was built with ethical rules from the beginning. My sources say it includes ideas such as “do no give financial advice”, “do NOT discuss race or discriminate” (or “do you not give medical advice”), and that OpenAI (the company building it) is currently working on it.
What I am trying to say is that “conversational language” is essential, but some very intelligent people (I won’t name names) are really just jerks. Chatbots such as ChatGPT require refined, deep content in order to build industrial-strength intelligence. If you are using the chatbot to overcome writer’s block, it is fine if it works “pretty well”. If you want the chatbot to work reliably, it must source valid, deep, extensive domain data.
One example is Elon Musk’s highly-hyped automated driving software. One, I don’t want any of the 99% unsafe cars on my road. It’s not enough to be 99.9% safe. This is also true: If the information corpus is flawed or the algorithms don’t “constantly verify for reliability,” ChatGPT could be a disinformation machine. Simply because it consumes so much data, it will likely be biased.
Imagine that the Russians could use GPT-3 to create a chatbot about the United States Government Policy and point it to every conspiracy theory website. This seems easy enough, and many people will use it if they place an American Flag on it. It is crucial to know where the information came from.
AI engineers understand this and believe that more data is better. OpenAI CEO Sam Altman believes that these systems can “learn from invalid data” as long as there is more data. Although I can understand the idea, I believe that it is false. ChatGPT OpenAI will have the greatest business benefits if it is used to point at trusted, smaller, validated deep databases.
The most impressive demos that I have seen over the years are those that focus on one domain. Paradox’s AI chatbot Olivia is intelligent enough to screen, interview and hire McDonald’s employees with remarkable effectiveness. A chatbot that manages bank compliance is available from a vendor. It works well and acts as a “chief Compliance Officer”.
As discusses in the podcast imagine if an AI could point to all of our HR research and professional growth. It could even be smarter than me and be called a virtual Josh Bersin. This is what we are currently prototyping.
Last week, I was able to see a demonstration of a system that automatically created quizzes from existing courseware in data science and software engineering. It also included course outline and course objectives, as well as course outline and course outline templates. This type of work requires a lot cognitive effort from instructional designers and subject matter specialists. We can instantly release our content to the entire world by “pointing” the ChatGPT AI at it. We, as designers or experts, can train it behind-the scenes.
Think of the many applications that AI can make in business. These include onboarding, sales training and compliance training. Leadership development and personal and professional coaching. ChatGPT AI can solve the problem of “expertise delivery” if it is focused on trusted content domains (most companies have plenty of these).
Where will this market go? ChatGPT
The pioneers are often left with arrows in the back, as with all new technologies. ChatGPT may seem miraculous but we can predict that innovators will quickly advance, expand, and refine it. It’s possible to bet that most VC companies are writing blank checks to startups in this field, which means there is plenty of competition.
OpenAI and Microsoft will compete with other players (Google and Oracle, Salesforce, ServiceNow and Workday, etc.) and every major vendor will “bulk-up” in AI and machine learning expertise. ChatGPT OpenAI APIs can be integrated into Azure by Microsoft. This will allow thousands of developers to create domain-specific products and innovative solutions. It’s too early to know, but I believe that domain-specific and industry-specific solutions will prevail.
You can only imagine the “opportunity space” that is out there. There are many opportunities for leadership development, coaching and counseling, as well as technical training and customer service. This is why I believe that this market has great potential. (I tried to contact PayPal via their chatbot, but was so frustrated, I closed my account.
ChatGPT is similar to “mobile computing” in the early days. In those days, we considered it an “add-on”. It grew, expanded and matured over time. Today, most digital systems are designed for mobile first. They build entire tech stacks around it, and we study consumer behavior, markets and consumers through our phones. It will be the same here. Imagine what it will look like when your customers start asking questions about your products. This is a huge opportunity.
As I discussed in the podcast, many jobs will be affected. ChatGPT has a direct impact on all jobs (editors and reporters, analysts, customer service representatives, QA engineers, etc.). I have just done an analysis. With 10.3 million jobs available, I found that about 8% of them (800,000.) will be affected immediately. These jobs will not disappear, but these systems will continue to improve and enhance them over time. (There are many new jobs, such as “Chatbot trainer”, that are being created.
ChatGPT is far from overdue. I invite you join me as a Josh Bersin Academy, or Corporate member to learn more. We’re always interested in hearing about your experiences or if you are building something.
Let’s keep moving forward: We can think of this as one the brightest stars of our future and prevent it from spiraling out of control.
- Codeplayon Jetpack Compose Tutorial
- Codeplayon Android Tutorial
- Codeplayon Flutter Tutorial
- Codeplayon on Github