Most Important Skills for a Product Engineer at Graphistry

I'll use this to talk about the current state of the creative industry, especially with the advent of ChatGPT. There's a question of what the industry will look like now and for software engineering specifically.

Briefly, I believe it's about finding what you care about and trusting your innate intelligence to guide you. You might discover new interests or realize you don't know what you care about yet, and that's okay. You can stay in that space until a new opportunity arises.

The AI aspect is particularly interesting. ChatGPT can code at a beginner's level in computer science, handling algorithms effectively. However, it has its strengths and weaknesses; it's good at some things and not as much at others.

Currently, some people are experimenting with prompt systems, hoping to build entire applications by organizing prompts in specific ways. I'm skeptical about these methods. I've found that ChatGPT struggles with layers of meaning, even when attempting tasks like generating psychological stories.

Similarly, in coding, I've observed that AI achieves local coherence but lacks consistency across larger structures, like an entire system. People are using diffs in Git repositories, which represent changes across multiple files. I'm curious to see where this leads, as bots are emerging that can evolve codebases incrementally.

My job often involves noticing structural differences between files and unifying them to simplify and clarify the entire system. A larger context window in large language models isn't sufficient on its own.

The evolution of AI capabilities is tied to training data. ChatGPT performs well when trained on specific coding problems from sources like Stack Overflow and Wikipedia. However, it struggles with problems that haven't been publicized.

AI is clearly advancing, with uneven progress. It excels in areas like image generation, though limitations exist. The question remains how the AI industry's overall capabilities are evolving and how they will integrate into other industries.

There are opportunities to increase people's capacity. Artists might feel threatened, and it's unfortunate that it seems tied to capitalism. While it might be harder to be a graphic designer in some ways, it also becomes easier to be prolific. With more powerful tools, designers can focus less on minute details and more on larger projects, like building an entire e-commerce site.

I see this as an increase in human capacity, and approaching it from that perspective allows one to benefit from the growth. I tend not to be political about these developments, believing that much of what's coming with AI capabilities cannot be changed.

Whether AI will reach human-level intelligence is uncertain, but it's unlikely to happen soon, certainly not within five years, and potentially much longer. I believe training data will be the primary bottleneck. Prompts are exciting for influencers, but actual engineering still relies on systems design and real-world experience.