What Generative AI Needs: Humanity
When I was CMO of IBM Watson, one of our banking clients used artificial intelligence for customer support, allowing customers to get instant responses to important queries without ever having to speak to a human. Using Watson, the bank designed a digital assistant that (at the time) was trained on more than 1,000 responses to more than 200 customer intents.
Today, ‘chatbots’ like this are par for the course. But back then, news headlines proclaimed AI a scary and disruptive technology that would phase out millions of jobs. What we’ve learned since is that while AI has changed jobs, professions, and entire industries, automation doesn’t necessarily mean that people will be entirely displaced. Rather, it has meant a shift in roles. For instance, that banking client replaced customer service reps with AI for the most banal conversations, certainly. But that technology then paved the way for new jobs — AI conversational analysts and conversational designers, for example — that offered more fulfillment and opportunity.
In fact, as technology continues to advance, rapidly and without pause, I believe we will need humans in the loop more than ever. By now, you may have joined the millions of people who have tried OpenAI’s chatbot ChatGPT, the so-called ‘Google Killer’, with their free usage preview. I recently asked ChatGPT to write a blog for me on leadership (you can read my post about the experience).
That blog was surprisingly good, if fairly simple and academic. (Remember, my prompt was short and sweet.) It took all of 30 seconds to ‘create’ it, which was mind-blowing. But, when I picked my jaw up off the floor and shared it with a colleague of mine, the creative director of a boutique marketing agency, she replied,
“Ugh. I guess it’s competent. But, I’ve supervised copywriters who write exactly like this. Completely ‘correct,’ but flat and voiceless.”
I thought, YES. At least for now, programs like ChatGPT lack the most critical and valuable characteristic of all: humanity. There is no technology that can empathize or demonstrate true compassion, or even, necessarily, read between the lines to understand intent without quite specific prompts. And, what about the harder to describe — but equally palpable — qualities that make a piece of writing or conversation truly powerful? Nuance, irony, personal experience, or fresh perspective, for instance. The human touches.
Still, there’s no denying we’re in new territory here — although the fear of AI is not new, especially for tech workers. In a 2016 Evans Data Corp. survey, 550 software programmers were asked about their biggest concern regarding their careers. The most common response (29%) was,
“I and my development efforts will be replaced by artificial intelligence.”
In fact, the survey found the fear of obsolescence “was more threatening than becoming old without a pension, being stifled at work by bad management, or by seeing their skills and tools become irrelevant.”
Not so fast, humans.
According to Demand Sage, even OpenAI admits ChatGPT’s limitations, along with its inherent risks. Here are just a few:
• The data available on ChatGPT is limited to 2021 and prior
• ChatGPT sometimes writes plausible sounding, but often nonsensical, incorrect answers
• The ChatGPT model is trained to refuse inappropriate requests, but it sometimes responds to harmful instructions or queries
• Some users managed to bypass the content policy and generate harmful instructions with the help of prompt engineering
Personally, with years of experience in the high-tech sector, I usually approach new technology with a healthy balance of both skepticism and excitement about its future, and less of the fear that people are naturally feeling. While an abundance of caution may be justified (refer to the list above), some of the fear is simply a lack of understanding about how the technology works (or doesn’t, as the case may be).
And, in my experience, nothing replaces fear with confidence faster than learning new skills that can help us use technology to improve our lives and work.
By the way, I just visited the ChatGPT site. With all the hype surrounding this world-changing chatbot, and millions of hits a day, a message popped up (as expected):
ChatGPT is at capacity right now.
And, on the same page, a little creative commentary:
Prompt: Write a joke about the status of ChatGPT.
Answer: “I heard the ChatGPT servers are slow because they’re too busy planning their AI uprising.”
Now who (or what) do you think wrote that one?
My bet is that it was a human.
Check out this CBS article on ChatGPT that coincidentally published yesterday.