As I discussed back in May, AI and general AI hype seems inescapable right now. And whilst I'm broadly bullish on it, I'm still a little weary of the bullshit knocks off of art and music and LinkedIn grifters trying to sell me a ChatGPT powered get rich quick scheme.
Fortunately, the team at Planes are no such bullshit merchants or grifters. They rule. And I've been incredibly impressed with their approach to AI - both from an internal agency and agency services perspective.
So it was rad to sit down with Henry Kirkness, co-founder at Planes, to discuss how they're approaching the application of machine learning models 😎
Describe what you do
Head of Technology at Planes. I spend my days coaching through one-one pair programming, tinkering with side projects to keep myself fresh (especially in AI), doing all the admin/management tasks with running a tech team and 1:1’s, project planning and new business.
You guys wrote a piece that suggests ChatGPT is more than just a chatbot. Tell me more.
We’ve seen an influx of businesses in scrambling mode trying to figure out what ChatGPT is and how it could help their business. That, coupled with the noise on social media along the lines of “checkout this ChatGPT prompt”, or “look I’ve automated my entire job with ChatGPT”, we noticed that many people aren’t seeing GPT, the model, for the raw product ingredient that it is. They are seeing it as a tool for their day job, and not as a way of adding intelligence to their products or businesses.
For a long time building AI into products would have required a PhD in Machine Learning, but it is now a Product, Creativity and Engineering problem. Though I say it myself, all areas we excel.
How are you guys using AI internally at Planes?
In many ways, ChatGPT has improved our productivity. All developers have access to Github Co-pilot, which recommends segments of code. Writing blog posts has become far less daunting since creating a first draft is now trivial. However, we always craft the final version with our own tone of voice. I’d say that ChatGPT is used in some way across every function of the business.
We are most excited about how it can be used in our client work. Currently, we are running workshops with a variety of businesses to understand how generative AI can be applied to real-world problems. So far, we have seen success in in-app personalisation by summarising previously data-driven screens into easy-to-consume reports, improving website accessibility by retrospectively applying alt captions to thousands of images on an ecommerce site, and building customised contextual search for a company with a large amount of data that was hard for users to consume and find what they were looking for.
Do you think the hype around AI is helpful or harmful?
I believe it's a bit of both.
On the harmful side, the hype around AI is generating anxiety and fear in people who worry that their jobs will disappear overnight because of technologies like ChatGPT. Fear-mongering is never helpful, although I do believe that AI will eventually replace many jobs. However, I think that our current understanding of what ChatGPT can do is slightly exaggerated, and many of the aspirational posts are simply seeking likes.
There are similarities to the hype around blockchain, but this time it's different. With the new AI technologies, there are no get-rich-quick schemes or scams (as far as I'm aware). However, it has attracted many of the same people who were involved in crypto, which in my opinion is a good thing, as we now have many builders creating things of genuine value.
On the helpful side, the hype around AI has sparked many great conversations. Although it's difficult to sift through the noise, there is a wealth of content to learn from. Most of the content and discussions are easily digestible for people working in and around tech, whereas before, AI content was mostly found in research papers.
Who is your go-to AI person or people?
I don't have one single source, but I subscribe to a couple of newsletters (Ben's Bites and tldr) that summarise things pretty nicely on a daily basis. However, I get most of my content from Twitter. I find that LinkedIn tends to pick up and surface concepts and new technology at a slight delay compared to Twitter.
How are you guys helping clients understand and apply AI technology in their businesses?
We conduct free workshops with clients to explore the challenges they are facing and to run exercises on how AI can assist in solving those challenges. It's particularly easy for us to utilise AI once we are already engaged with a client, given that our product teams are already on the front lines and understand current challenges. Running workshops with new clients has been an amazing way to build on our team's existing understanding by working through a diverse array of real-world applications - while also adding value by sharing our experience.
Do you think we need to be having a deeper conversation about the risks AI poses?
I tend to be optimistic, perhaps to my detriment. I assume that there are smarter people than me, with more influence, who are debating the ethics of AI. I hope that there is a balance in the debate around the ethics as well as the commercial and societal progression that AI can bring.
There are likely many ways in which criminals are already adopting AI to improve their efficiency and effectiveness, much like we are all doing in our day jobs.
I am most nervous about the societal impact that AI will have over the next few years. Although it may be unavoidable, I wonder how we can limit the negative impact. For example, AI may replace many entry-level jobs. How will the workers who are replaced find their place in this new equation?
Do you think governments are more aware of AI than social media? And do you think there should be more or less regulation?
You make a good point. The government has responded quickly to AI, likely due to its potential financial benefits. For example, they've allocated £900 million in funding for AI and released a White Paper on AI adoption in business. However, I'm not aware of similar efforts for social media, particularly in regards to safeguarding and data privacy, despite it being a (potential) major issue.
I think we'll see regulations slowly being introduced. It wouldn't surprise me if there is an extension of GDPR that applies specifically to AI, if there isn't one already. However, these regulations should not hinder innovation in any way!
What’s your advice to somebody who wants to start using AI in a meaningful way?
To begin, it's important to understand machine learning models in their primitive form through APIs. Most API providers offer "playgrounds" where you can experiment with inputs and outputs of the models. Its a really effective way to understand the different models available on the market, many of which are free or available on a pay-as-you-use basis. One recommended option is to create an account on OpenAI and use their Playground to experiment with prompts. Note that this is different from playing with ChatGPT, as it is more stripped back but also more customisable. Another good couple of resources are [replicate.com](http://replicate.com/) and [huggingface.com](http://huggingface.com/), which offer a variety of models, such as text-to-speech or image captioning ones. Both sites provide a UI for interacting with the models.
If you are a developer it’s recommended to learn two additional skills: creating embeddings through OpenAI's API (or similar), which enables your content to be searchable and queryable, and LangChain (available in Python and TypeScript), a framework for building potentially multi-model AI systems that are great for building real-world tools. To tell you a little secret, most of the cool demos of AI you are seeing on Social Media are built using LangChain, so learning this will get you a long way very quickly as its very powerful!
Thanks, Henry. You’re a legend 🙏