I grew up in a very religious family. I don't view this fact as bad or good, but it's important to mention as part of the story I am telling in this blog post. As a teenager, most of my Friday nights were spent at youth group, volunteering to pray for people or tidying up at the end of the night. During that time I was recruited to help lead a group called Promised Girls (yes, you can laugh because its exactly what you think it is!) The goal of this girls only group was the pursuit of purity in service to god and to our future husbands. To do things the right way. A central component of this group was meeting to discuss how terrible it was that other girls were dating around while we, the pure, were gathered together praying for our husbands (that we had not met yet.) Sometimes we'd have guest speakers come in to tell us that the reason the other girls were dating in high school was because they were broken in some way, or had low self esteem or something like that. When I think back, part of me is proud of the person I was. Strong willed, not easily influenced, morally certain. I think this is a good place to be but has many downsides.
The other day I was scrolling on LinkedIn and saw a blog post by a good friend of mine from the Python community. The blogpost was discussing why any ethical person would be working on AI at a Big tech company or in one of the labs. This is a fair question to ask. Someone in the comments said he thinks it's mainly people from third world countries that don't have a choice. Another said its people who are too tied to the money, and the right way to do things is to work at a company you trust. This conversation has been bouncing around in my head for two weeks now and I thought I'd explore it here.
I recently watched Dario and Daniela Amodei from Anthropic being interviewed by Oprah. When asked why they were racing to build a technology that Dario has said could cause human extinction, Dario responded by saying,
'''
This technology is an epochal change for humanity. At some point humans started to use fire. At some point humans built steam engines and factories, and that created the Industrial Revolution... But we understand that the Industrial Revolution as a process, as a technological event, was worth doing. We don't want to be living in caves, but we need to manage it in the right way.
The way we think about this company is: there are many companies building this technology. It's something that is happening to humanity one way or another. We're in it. There's this train going very fast in some direction. You don't want it to crash. You can't stop the train. But what you can do is steer the train so that it doesn't hit the rocks. The way we think about Anthropic is we want to do things the right way. We want to decrease that risk, hopefully to zero, and do things the right way.
'''
In this one response Dario mentioned doing things 'the right way' 3 times!!! What does it actually mean to do this??? Whats being implied is that AI is already here and it is a good thing, and is not going to stop, but the train needs brakes and people need to be warned about it. At the Code with Claude conference in London a few days ago, I had the chance to speak to some Anthropic employees. I asked them what they thought. The conversation drifted to Mythos and Project Glasswing and why for some things it was important to get there first, so you could ensure safety. I thought this was fair, I am not sure if it is sufficient. Many people online have said they think the ethical positioning that is done by Anthropic is fear mongering or a marketing ploy. Again, I'm not really sure about this.
On the opposite side of the spectrum. A few weeks ago Anil Dash, former founder of Glitch and former director at StackOverflow, shared a blog post he titled '(One) Good AI is here'. In his LinkedIn post about it he wrote,
'''
Those of us who share the valid criticisms of Big AI, but who also believe that machine learning systems could be helpful if made more accountable, have been waiting for tools that are made the right way. I've been saying it's possible to have an AI that only uses consensually-gathered content, is open source + open weights, runs on your own machine, and is designed to empower creators instead of exploit them...
'''
You should read the blogpost yourself but Anil highlights a video-generation AI tool created by Corridor Digital. The tool helps creators turn videos into stylised visuals while keeping the original acting, motion, and camera work. They made a video about it here that has 1.9million views! Though the same techniques were used to build it, the difference between CorridorKey and Sora or VEO, is the way it was built and is being distributed. Ethically sourced data, open weight and free to use. Though we haven't met, Anil is someone I have a lot of respect for, and he says this is the right way to do things.
I personally love open source and think a shining light about the Python community is how thoughtful they are about ethics and how technology affects people. My doubts around this are that open weight models and AI techniques all somewhat rely on research coming out of the big labs. I am not sure if we can ever purify ourselves from the process it took to get here. It's a worthy pursuit to critique and rethink the way we build these systems. I see open weight models becoming a big part of AI systems in the future, but I am hesitant to say that this is the correct way to do all things. I have lots of thoughts on this but I'll save them for a longer blog post in the future.
Does purity matter and what is the right way to do AI?
I am no longer a teenager anymore and I no longer believe that girls in my high school went on dates because there was something wrong with them. In hindsight I think going on some dates would have done me some good back then so I could make an informed decision for myself on the subject instead of relying on others to tell me what to do. This is also the conclusion I have come to on the subject of how to use/do AI, at least so far.
People are using AI very differently. At AI Engineer Europe, my friend Mario Zechner, who created Pi (allegedly Armin Ronacher created it too), spoke about how we need to slow down, maintain human oversight and try to be thoughtful about which tasks we allocate to agents and where humans should get involved with the code. His talk was contrasted by one from OpenAI's Ryan Lopopolo, who discussed how code is an abundant resource and that we should allow agents to complete full engineering tasks with humans steering the process but not writing any code themselves. Immediately after this conference I spoke at DjangoCon Europe and met two engineers who were not using AI in their workflows at all. They didn't really have an ethical or moral reason for this, one was afraid of it making mistakes and the other just mentioned it was too much of a cognitive load to learn something new. Several engineers I know work at companies that have fired people because of AI. At the same time many engineers are having their AI coding spend capped because of the cost of tokens. It's a lot going on!!!
I want to end this blogpost by saying the best thing I believe you can do for yourself is to listen and to experiment. Depending on who you're following on LinkedIn or which conference you attend or which company you work for, you'll probably have a different thing in your mind when someone talks about the right way to do AI. The spectrum is wide and I think its helpful to listen to as many perspectives as you can so you know this. Ultimately, you are the engineer that will need to survive the future of this industry regardless of what people are writing in their blogs. You are also a human that will probably need to reconcile what you believe and where you stand ethically and how this shapes your actions. I am still grappling with these thoughts myself, and would love to hear your thoughts if you have some. Feel free to leave me a message if you'd like. Thanks for reading!
💬 comments