A diverse background (and a new opportunity)

Simon is a U.K.-based Senior QA with a Bachelor and Master’s in teaching. He spent ten years working as an English teacher, both internationally and in England. After a career change, Simon transitioned from freelance editing to working as a managing editor for a blockchain company. 

Mindrift Team: How did you first engage with AI projects at Mindrift?

Simon: When I first joined Mindrift, I had no knowledge of chatbot development or AI processes. Initially, I worked on simple tasks like evaluating video content or generating chatbot responses. Over time, as projects became more domain-specific, I found them more aligned with my interests.

Mindrift Team: What aspects of AI interested you? Was it a natural transition from blockchain?

Simon: My main interest is still education. In the UK, teacher dropout rates are at an all-time high due to heavy administrative burdens on new teachers. AI could help automate lesson planning, which takes up a huge portion of teachers' workload. LLMs can analyze classroom profiles, goals, and subjects to create structured plans, making teachers' lives easier. AI can also improve virtual learning environments.

The nitty-gritty of AI training

After collaborating with Mindrift for over a year, Simon has seen first-hand how projects and expectations shift due to industry changes. 

Mindrift Team: Have projects become more complex over the last year?

Simon: Yes, across the industry. Early AI projects focused on training base-level LLMs, but the clients have moved on to building niche, highly specialized AI models. Some projects involve complex research and require high-level domain expertise. This shift has changed freelancers recruitment processes, project setup, and quality assurance, with tasks now taking significantly longer to complete. Previously, a QA could check tasks in minutes—now, that might take hours.

Mindrift Team: So your main role is ensuring alignment with increasingly complex guidelines?

Simon: Exactly. Client demands are much more refined, which means our work is more intricate. Some guidelines are 60 pages long, and tasks that once took minutes now take hours.

Mindrift Team: What makes a good prompt?

Simon: It varies by project, but generally, a good prompt feels natural. Many writers over-structure their prompts, making them read like instruction lists instead of realistic dialogue. A good prompt should reflect a real-world persona and be both natural and strictly compliant with client guidelines.

Mindrift Team: Are editors involved in refining prompts?

Simon: Editing is shifting from a traditional phase to more of a content discussion. Some tasks take 10 hours to write and 8 hours to review. Editors now focus more on technical compliance, leaving much of the writing responsibility to the writers.

Looking to the future of AI

Since our goal is to advance ethical AI models, we love to hear our freelancers’ hot takes and opinions on AI—past, present, and future. 

Mindrift Team: What surprised you about AI projects?

Simon: Initially, I was surprised by how LLMs learn—essentially by throwing vast amounts of data at them. Early projects involved creating thousands of repetitive tasks, and I expected the process to be more complex. Now, as tasks become more intricate, my understanding has deepened.

Mindrift Team: AI still relies heavily on human input, despite advancements.

Simon: Yes, and I think we’re in a phase of rediscovering the need for human refinement. Early LLMs relied on raw big data, but quality suffered. Now, human oversight is more important than ever.

Mindrift Team: Can you still distinguish AI-generated text from human writing?

Simon: Easily. AI tends to overuse transition words like "however" and "additionally." It struggles with flexible, natural phrasing and slang, making its output formulaic.

Mindrift Team: So AI is more suited for structuring and organizing data?

Simon: Yes, it’s excellent for generating a starting point—like research summaries or lesson plan outlines. But human input is still essential for refining and personalizing the output. Over time, AI will likely contribute more, but humans will remain crucial in shaping quality content.

Mindrift Team: Some see AI as a threat. Do you?

Simon: Not really. Every new technology is met with skepticism. The internet was once seen as a passing fad that would eliminate jobs, yet it ultimately created as many opportunities as it displaced. AI will likely follow the same path—it will reshape industries but won’t make human creativity obsolete.

Mindrift Team: Market disruption is a huge topic, and AI is at the center of it.

Simon: Exactly. My editing business nearly collapsed overnight when AI tools became widespread. Clients stopped coming to me for editing, thinking AI could replace it. But within a year, they returned, realizing AI-produced work needed human refinement. It’s a classic case of market disruption, where we first overestimate a technology’s capabilities before finding its proper role.

Critical skills required

With AI projects becoming more niche and complex, Mindrift created the talent pool concept—a community of domain experts who want to contribute their expertise to advancing AI models. 

Mindrift Team: How do domain experts adapt to AI tasks?

Simon: One challenge is that we now recruit domain experts, but expertise in a field doesn’t automatically translate to success in AI tasks. For example, a PhD in medicine doesn’t mean someone can craft effective LLM prompts. This disconnect leads to frustration when experts struggle with the tasks and fail to meet requirements. We need to improve how we communicate these expectations during recruitment.

Mindrift Team: So writing skills and adherence to guidelines are really important, right? 

Simon: Absolutely. Experts often skim guidelines because they don't see them as important, but in reality, writing quality and strict compliance with instructions are critical.

If this sounds like you, and you have professional experience in any domain, come join us at Mindrift! Mindrift is a platform that curates a talent pool of experts and connects them with cutting-edge AI projects from the world’s leading companies. 

Work on interesting AI projects on your own time, add new experiences to your resume, and get paid for your contribution. Check out our open roles and talent pools to get started.