The AI Cynic Who Released an AI Product
Daniel Heuman, CEO and Founder, Intelligent Editing
I grew up reading Isaac Asimov and watching Star Trek: The Next Generation. The similarity between my name and Asimov’s “Daneel Hummin” is 100% coincidental (as is my physical resemblance to Spock). However, it’s fair to say that I’ve been interested in AI for a long time!
What about AI editing suggestions? No thank you! I've always been sceptical of AI editing. It inundates you with useless recommendations and insufficient explanation. It’s hard to see why that would be helpful.
So why would an AI cynic, like myself, create Draftsmith? An AI provides all of Draftsmith’s suggestions. And, as with any AI, it produces many useless recommendations!
AI Isn’t Intelligent
It helps to understand the science fiction I grew up on and how it relates to what is available today. Today's AI may be impressive to us, but it's not yet close to becoming the sentient beings of sci-fi lore. Noam Chomsky’s takedown of ChatGPT superbly sums it up:
“Their deepest flaw is the absence of the most critical capacity of any intelligence: to say not only what is the case, what was the case and what will be the case — that’s description and prediction — but also what is not the case and what could and could not be the case. Those are the ingredients of explanation, the mark of true intelligence.”
In other words, ChatGPT lacks true intelligence. GPT 4 lacks true intelligence. No AI model that anyone has produced to date has true intelligence. And that’s my view before we even get into the awful problems of bias and ethics. The material that you use to teach an AI is crucial for its learning process. And it often leads to surprising, mysterious or completely wrong results.
So why release an AI product I helped design?
People Make the Best Editing Decisions
Above all, I want Intelligent Editing to build products for language professionals who work with words. I care deeply about the craft of editing and the importance of editors. This technology exists now. And even though it makes some bad suggestions, it has advanced to a point where it can help professionals work and interact with text. The helpful suggestions, and the efficiency that they offer, vastly outweigh the bad suggestions (that are easy to ignore).
After years of tracking AI developments (and shaking my head at every turn), why now? For me, Harriet Power’s article for CIEP on ChatGPT versus a Human Editor was a turning point. She found that “ChatGPT caught 15 out of 17 errors” on CIEP’s proofreading test. If you were waiting for evidence on what editors want and need, it’s hard to imagine a clearer statement.
That makes it sound like AI is on the way to replacing human editors. However, I don’t think that will ever happen.
For all that AI editing seems magical, it can’t reliably look things up. It can’t manage the people involved. And, most importantly, it lacks the nuance of a human editor. It has no respect for the author’s voice. So what author would choose that?
That’s why our AI product is designed for earlier in the document creation process. Editors may find it useful in certain situations (especially when dealing with text that requires an extensive edit). But Draftsmith is really designed for writing refinement.
Writing refinements consists of the reviewing, revising and redrafting phase of document creation. That’s where authors can be inspired by AI suggestions, and where they can easily skip past useless suggestions. Writing refinement starts after the first draft is finished, and it ends when the work is ready for a final polish and copyedit.
Innovation, Ethics and Security
As CEO of Intelligent Editing, I'm responsible for ensuring our products are innovative and ethical. Draftsmith meets both those criteria. It uses AI in new ways. It doesn’t replace editors. I think it supports our community of language professionals who work with words. I wouldn’t have allowed it to progress if it didn’t.
On security, Draftsmith uses Microsoft’s Azure OpenAI service. OpenAI are the makers of ChatGPT. So if you’re allowed to use ChatGPT at work then you can probably use Draftsmith. If not, then you probably can’t.
We’ve made Draftsmith as secure as possible. You choose what you send (sentence by sentence). So nothing leaves your computer without your permission. Moreover, Microsoft’s Azure OpenAI service has enterprise-grade security. Your data is protected by comprehensive enterprise compliance and security controls. Your data is not used to train any foundational AI models. We’re opting out of Microsoft’s human review. And as a company, we don’t see your data either. No one at Intelligent Editing has access to your text. We store only the metadata (the size of what you sent, when you sent it, and what prompt is applied).
My Cynicism About AI Editing Hasn’t Changed
I’m still an AI cynic. Everyone should continue to push for transparency, fairness, and ethical practices in the development and use of AI. We must also be mindful of the potential social and economic impacts of AI, and work to address all of the popular misconceptions about what it is. From there, we can work to stop its misuse.
I’ll always be an AI cynic when it comes to editing, but the technology has changed. If you’re a professional who works with words, and you want to know when to start considering AI at work, I can tell you what I’m doing. I wrote this article with (a little) help from ChatGPT. Then I redrafted it with some help from Draftsmith. It will go for a human edit (and a check with PerfectIt) before it’s published because that part can’t be replaced. If you ask me, I think the time to start considering where AI can help in your work is now.