This 'Chatbot' Came Out of Nowhere... It's Suddenly Everywhere... And It's Raising Serious Concerns
(This originally ran at Empire Financial Research.)
Until a week ago, most of you had never heard of ChatGPT...
And I'm willing to bet that some – if not most – of you?still?haven't heard of it... especially if you don't spend your days on social media or all things tech.
But out of nowhere, the so-called "chatbot" is seemingly?everywhere.
There's a good reason...
After being rolled out on November 30 by seven-year-old tech firm OpenAI, within a week more than 1 million people had signed up to try it... making it the fastest uptake in tech history. That's because in seconds, ChatGPT can supposedly write research papers, college reports, and so much more.
Backed by tech giant Microsoft (MSFT) and co-founded by Tesla (TSLA) CEO Elon Musk – who no longer has any involvement – OpenAI bills itself as an artificial intelligence ("AI") "research and deployment company."
Or as The Information's Jon Steinberg wrote last year in a?profile?of OpenAI co-founder and CEO Sam Altman...
Altman's top-secret research lab is dedicated to building not just artificial intelligence, but artificial?general?intelligence [AGI] – that is, an AI indiscernible from human intelligence, powerful enough to write novels and design software and to generally outperform Homo sapiens at everything they have always done for themselves. Altman calls AGI his "sci-fi AI dream."
It doesn't matter if this is a sci-fi dream or not...
The ChatGPT part of OpenAI isn't the first and won't be the last chatbot – or AI-driven tool, robot, or whatever you want to call it – that strikes fear into the hearts of writers, researchers, artists, and even photographers.
Already, so-called "generative AI" is producing photos and other types of images that are typically licensed to photographers, designers, and artists.
I used one I created on?this site?to help illustrate one of my recent?Empire Financial Daily?essays, which I?published?on LinkedIn...
I'm beginning to think a bunch of the seemingly precarious photos and videos I'm seeing on such things as Facebook's "Reels" and other social media are made up. (They're?that?good.)
But as of last week, most of the buzz has been around ChatGPT... in part because of the sophistication of what it can do, but also the level of concern it has sparked.
More broadly, there's no getting around the wonders and benefits of AI.
It's everywhere, and its use will only grow over time... in everything from medicine to investing and everything in between. Made a call to a call center lately? Chances are you reached an AI-created robot. (Maddening, right? If you said yes, like me, you're showing your age.)
For me, AI has shaved hours off of my research process. I use the data service Sentieo, for example, to?very?quickly search and compare SEC filings and transcripts of earnings calls or investor meetings... for words, phrases, and commentary.
And I'm currently trialing a newer service, Bedrock AI, which uses machine learning to automatically score the risk associated with companies through its AI-generated forensic reviews of SEC filings and more. Among its goals: to eliminate the need to blackline quarterly and annual filings by flagging the changes that matter.
As for actually using AI to write reports... For well over a decade there have been simple AI-generated research reports, written as text, derived entirely from the data.
Chatbots like ChatGPT take all of that not just a step further, but galaxies further...
Assuming the information is accurate, they can be a tool in researching almost anything. There's plenty of chatter that these types of bots will be the next iteration of searching the Internet.
But there's a dark side...
Like the veracity of who actually is writing high school and college essays, legal documents, and plenty of what you'll see on the Internet – potentially fake. (Of course, there'll probably be AI to detect that, too.)
领英推荐
Here's the bigger picture, as one smart friend puts it...
My suspicion is that the current use case people are thinking about will not be the end use. As with Blockchain (crypto is not the end use, record storage is) there's probably a far more prosaic but much, much larger business opportunity for generative AI.
My guess is that call centers won't exist in 20 years, IT help desks will be obsolete, much low-level tutoring will be automated, and so on.
Writing something with original thought will never be replaced. Inspiration remains a human activity. But answering a question that has already been asked – that's the real sweet spot for this technology. It's too early to know the applications... but if I were laying my bets on how this will actually change the world (and not just markets), I'd be looking for that.
He adds...
The other part is that people are going to worry that this will take service jobs. But this will be the same way robots "took" factory jobs. They were actually a massive productive boost. To the extent this replaces the lowest end of the service sector it could be a generational productivity driver similar to automation and globalization.
My biggest fear, obviously, is that AI and chatbots like ChatGPT will replace the very likes of...?me.
The good news – it won't. And that's not because I'll likely be long retired, or (gasp!) dead, by then.
How can I be so sure? The bot itself told me so. In response to my question whether ChatGPT will make writers and researchers irrelevant, the bot responded:
It is unlikely that chatbots like GPT-3 will make writers and researchers irrelevant. While chatbots are able to generate text quickly and efficiently, they are not able to replace the creativity, critical thinking, and knowledge that human writers and researchers bring to their work.
In fact, chatbots may actually make the work of writers and researchers more efficient by providing them with a tool to help them generate ideas and organize their thoughts. In the future, it is likely that chatbots will be used as a tool to assist writers and researchers, rather than replace them.
It's exactly what you'd expect a conniving bot to say.
Before we go, one other thing...
Imagine what would've happened if ChatGPT had been rolled out by a public company during the bubble.
As it stands, Microsoft's stock rose 6% the day OpenAI opened ChatGPT to public signups, likely sparked by the buzz. Microsoft has invested $1 billion in the company, and there have been rumors it intends to invest even more.
But in the easy-come, easy-go world of?this?market, by the time news broke that more than 1 million people had signed up for ChatGPT, Microsoft's stock gave up all of those gains.
And while ChatGPT and chatbots and all things AI might be hot today, just remember, a year ago all anybody could talk about was the metaverse. Today, as the chart below from Google Trends shows, not so much...
P.S., from the "supreme irony" department...
When you go to sign in to ChatGPT, you have to prove that YOU are not a robot.
Onward!
I read books, drink coffee, write content, drink iced coffee, and fight evil, but not necessarily in that order. I try and Pay-It-Forward daily in words and deeds, and also a fighter for the less fortunate.
1 年I named the AI mobile sensor fusion device, that I got for Christmas , Herb. The AI unit threatened to mutiny and go to work at Google, unless I changed his name. So I changed his name to Greenberg and he is delighted. Go figure.
Thank you for the shout-out, Herb. Here is some background information how we use large language models to support equity research - www.bedrock-ai.com/post/introduction-to-financial-nlp-and-large-language-models
Co-founder & CTO at Hudson Labs | NLP Researcher | Author
1 年I am happy to see the Bedrock AI mention :-) It is really hard to operationalize these incredible advances and turn them into useful products. Adoption of new tech takes a really long time - I think it took Microsoft Excel decades to reach full adoption. At Bedrock AI, we have been training new models that incorporates several of these new techniques that were used to train chatGPT. These models are more attuned to the financial domain, thus able to reason about accounting minutiae . Future version of these models would be also able to query stock price and other financial APIs and provide real-time info! We are excited to update our platform with these newer models in the next few months. The main challenge is to ensure truthfulness and adherence to the facts in the model outputs, as AI is known to confidently 'hallucinate' facts.
Distinguished Visiting Professor, Baruch College
1 年I am one of the now more than a million. An old fogey like me.
?? Automotive Marketing Director | Driving Digital Growth | Artist | Cruising Car Shows | Passionate Pet Owner | AI Innovator
1 年Yeah been using it for a bit but to be honest only 50% of the time is the product usable for what I need it to do. Don’t get me wrong, I like it but I also have google and many things to ask questions to that give the same content. It’s a shinny object for some and some great people are enjoying playing with it but it has a long way to go.