Making sense of generative AI
Interview with Yuying Chen-Wynn, AI Strategy & Product Advisor @ Wittingly Ventures
How should we think about ChatGPT?
Last November generative AI was considered just another new technology.
It was cool. Had a lot of promise. But frankly, it was no different than blockchain or quantum computing.
Fast forward just a few months and generative AI has captured the hearts and minds of the world. With the release of ChatGPT, Dall-E, Midjourney, Bing, and Google Bard, you can now talk to a computer like it is another human being.
The experience is surreal. It's a whole new interface. And it is all happening so fast.
So how do we make sense of everything that has been going on for the last few months? To help us decipher that, I interviewed Yuying Chen-Wynn.
Yuying Chen-Wynn is an AI Strategy & Product advisor at Wittingly Ventures. She is a tech executive with a passion for leveraging AI and data to create 10x user experiences throughout global markets in EdTech and FinTech.
Prior to Wittingly, she was the Chief Product Officer for Barnes & Noble Education. And held various product leadership positions at Age of Learning and EF Education first. She is also the author of the AI Productization Profile series, which can be found on Linked In.
What follows below is a condensed and lightly edited version of our interview.
Yuying, let's start with the basics. How do you think about capabilities like ChatGPT or Midjourney? Why is it different from the other AI chatbots that have come before?
Yuying Chen-Wynn: Thanks Mustafa, good to be here, and looking forward to our talk.
Definitions are a good place to start.
The way I look at it is conversational AI vs. generative AI. Conversational AI is what we are all familiar with. These are the customer service chatbots we see all over the place. When you type in a question, and if it is a predefined question with slight variations, conversational AI will respond with a standard or pre-programmed answer. They are really just fancy search tools. The answers are in a database and it pulls it out for you.
Generative AI, on the other hand, has the ability to generate brand-new responses for you. Yes, it relies on the world wide web for its database. Buts its true power is stitching all that data together and creating new content that never existed before.
It is creating original content. And that is what makes this such a fascinating technology.
But this is not new, right? If you think about it this has been around since the 1950s. So what is different now?
YW: For me, it is a couple of things.
First, it is finally good enough. Talking to ChatGPT is like talking to a person. You sometimes forget it is a machine you are talking to. I mean, it is generating a response that is just like how you and I generate a response when asked a question.
Second, it’s not just texting anymore. Some tools now generate images. And pretty soon we will have audio and video too. The possibilities are endless.
And third, this is the best part. It is all open source and free. It's the most powerful tool that we have and everyone has access to it.
You are right. You can't argue with free (laughing). But more importantly, I second your remarks. The experience of talking to ChatGPT is surreal. I know it is a computer but it is so easy to forget.
Anyways, if this is where we are today. Where do you think we are going with this? What is possible with this new technology?
YW: Well we are just digesting what is happening now. And it is a lot.
But if I had to look at my crystal ball, I would say we are just scratching the surface of creativity. Think about it. What if you had a TV show that was generated just for you? Complete customization. You are not watching someone else's story. Rather the story is created for you.
The other use case that comes to mind is that of a babel fish. You know the one that translates everything instantaneously. Well with generative AI now you can. And so then think about the possibilities for travel, for business, for international relationships. Language does not become a barrier.
The other use case I would personally like to see…and I know this is a bit out there, is a complete replica of me. Like a mini-me that will take care of all the mundane tasks that I have to take care of. For example, creating a first draft of a deck or an email, or a report. The generative AI will just do it for me. And then I would come in and put in the finishing touches.
Wow…I love those examples. The one that comes to my mind is that with generative AI we may not be so reliant on screens. Right now we use screens to interface with a computer because that is the most efficient way.
But if you can have a conversation, talking will be your first mode of interaction. And for more complex things it would be talking and then screens.
YW: Just like Star Trek.
Yes exactly. I could just say - “computer please analyze the following data and create a short memo”
YW: I literally just watched this demo with Microsoft Office Copilot. This type of functionality is not far off.
It’s fun to speculate about the future. But let's get back to the present. Do you know of any organizations that are currently using this technology? If so, how?
YW: Right now most of the natural language models are being used for text generation and maybe a little bit of image generation.
But the text is where the action is at. And those are the use cases you and I previously talked about.
For example, there is customer service. Where generative AI can produce more human-like responses. Salesforce uses this for Customer 360. Where generative AI drafts the first response, and sends it to the agent, who reviews and sends it out. According to their data, 95% of responses are sent out without any editing.
The second is marketing. Where you can make marketing copy to help you sell or promote a new product and service. Actually, Jasper.ai is quite good at that already and I know quite a few folks are using it.
The third is training, especially sales training where interaction is so critical. Now you can role-play with natural language models and figure out the right script for your sales pitch. This is especially true with pharmaceutical and biosciences companies, where sales training is highly technical.
Have you seen any examples in coding? I mean GitHub co-pilot is taking off?
YW: Yes, GitHub Copilot is very popular with developers. But I think a bigger market is coding for non-developers. That maturity is not there yet. That would unlock a lot of creativity.
I am curious, as these organizations are adapting to this new technology, what have been some of the challenges? Or struggles?
YW: Well let's take the marketing example that I talked about. Because companies like Jasper.ai have been around for a few years now.
I know quite a few marketers using generative AI tools but are shy about sharing. They will use it to make the first draft, maybe even close to the final draft. But they will not let their work or colleagues know that they are using the tool.
Part of it is just cultural. They don't want to give the impression that they are not working, or that they are cheating of some sort. It is also a little bit about job protection.
At the other end of the spectrum is that these marketers now have to learn a new skill. How to ask effective questions. To generate good copy they need to write the right prompts. Generic prompts won't do. And so there are now some marketers who are sharing and teaching what those prompts look like to other marketers. How to structure them.
It’s like a whole new world out there (laughing).
I agree, it's a bit topsy-turvy. Suddenly we have to now learn how to ask the right questions to a chatbot. It’s not like just adding new words like you do with Google search.
Let's turn it around. Have you encountered any start-ups that are doing some cool stuff with large natural language models and/or generative AI?
YW: Well the most obvious ones are to help us with writing. There are a lot of start-ups that are focusing on specific use cases, skills, and industries. For example, Jasper.ai, Copy.ai, Lavender.ai, Gamma.app. The one I really like is Write With Laika. Where you can ask it to write a story based on specific prompts or you can start a story and the AI can help you write it.
Then there is finance - Surmount.ai. Where the generative AI will create and execute a trading strategy based on your hypothesis and risk tolerance. It can back-test it, run simulations, and even connect you to your broker to execute. Now suddenly I can be an algorithmic trader just like professional traders.
The other one I like is a crossover with VR - It's a USC collaboration with the VA (veterans administration), the veteran’s administration. Where you can use generative AI to train therapists to deliver better care. Just like in the sales training example above, therapists also need a lot of practical training. Which is very hard. Now you can.
The ones that deliver therapy directly to patients include Bravemind, Recover VR, and OxfordVR. Although I'm not sure these integrate with generative AI, but Oxford integrates conversational AI.
What about the productivity side? Have you encountered any start-ups in that area that you are interested in?
YW: Yes the one I like for customer services and chatbots is Botcopy.
We already talked about the GitHub co-pilot. And then there is Jasper.ai for marketing.
There are a lot more popping up almost every day. The field is constantly changing.
Yuying, one last question. For leaders who want to learn more about generative AI and what is happening in the field, do you have any recommendations?
YW: A few come to mind. Sequoia has done a good job mapping the industry. Their AI fund report is a good place to start.
If you want to listen to thought leaders, the recording from the AI conference run by Jasper.ai is a great place to start. It is the first conference focused only on generative AI and it is a great resource.
Well, let's not forget the awesome work that you are doing with your articles on Linked In. After all, that is how I found you.
YW: Thank you for that shout-out.
I have been interviewing interesting start-ups and talking about their vision, their product, and the impact they are trying to drive. I just published my 8th article in my series to highlight AI in different industries: Gaming - AI Experimentation and ML Live Ops. You can find the rest on my LinkedIn profile.
Yuying, this has been a blast. Thank you for being a guest. We have to continue our conversation. There is such an evolving field.
YW: Completely agree. Would love to be back. Thank you for having me. This has been a lot of fun.