Free Photography for Training AI

Have you ever wondered what it would be like to train AI and get paid for it? Well, there are several companies that hire independent contractors to perform work to train AI.

From the start, training AI may seem like a daunting task, but these platforms break up the work into manageable pieces.

Here’s a general overview of how training AI works.

A Customer or Client decides they want to use AI for their work. Or maybe invest in a large language model (LLM) to offer use to customers at various levels of pricing. The Customer contacts one of many AI Training Companies to discuss their needs.

The Training company then begins mapping out what will be needed to create an LLM to adapt to the customers’ needs. There is a lot of work that goes into creating the basic program. Then more work to set up training User Interfaces (UIs) to allow humans to perform the training.

Similar to how you would teach a child to communicate, think, comprehend and reason, this is what an LLM needs to be taught. Because a child can learn through various senses such as sight, sound, touch, taste, and smell, they can simply absorb information and the brain processes it.

However, an LLM has to have everything fed to it through electronic devices such as computers, cell phones or tablets. Each aspect of its learning has to be covered. The model needs to be taught restraint, so it doesn’t offend or insult people or cause harm by passing on damaging information like how to build a bomb or perform illegal activities.

The model also has to have safeguards against hackers who might have nefarious intentions to cause damage to its programming.

Then the model needs to learn how to reason, how to ensure accuracy and how to interact with humans.

So each of these jobs is split into manageable tasks. AI has to learn slowly. The most effective teaching is the type that can be used over and over and applied to several different tasks.
It’s the old adage, “Give a man a fish, you feed him for a day. Teach a man to fish, you feed him for a lifetime!”

Each task provides the model with a prompt (instructions) then the model provides one or more answers. The person who is working on the task must follow a set of guidelines (Usually called a Rubric) by which to judge the model response. For example, is the answer helpful, friendly or accurate? This rating will help the developers adjust the model parameters so it performs the way the client wants it to.

Let’s assume the Client wants a helpful chatbot to give people information about a certain area, the tasks might include asking for relevant information about a location. For example, “Where are the best seafood restaurants in Sample Town?” or, “I’m moving into my first home in Sample Town, what services provide trash pickup or is that handled by the town?”

By asking the model these types of questions and grading its responses the model can learn what it’s done well and what it needs to improve.
Each company has its own unique way of presenting work for ICs (independent contractors) and their UI will look different but the main training boils down to breaking up the training information into small parts and dividing the work among a wide variety of people. Each person working on AI is only a small part of the big picture.

When you apply to one of these companies your experience will be different from one company to the next. But the main work will be a combination of many minds to help train the customer’s model.

Most companies will hire enough contractors before bidding on a job. This ensures if they win the contract they can get started immediately and betters their chances of making the deadlines given by the Client/Customer.

This is why you will often see people complain that they got hired, and performed free training and assessments but have not worked. They have come in before the projects are started.

Once the company bids on a job, they will often have a trial run. They will have their contractors perform a sampling of the work to send to the Client for approval. If the Client is happy the project goes full force and more ICs are put on the project.

If you’re lucky to get on a large scale project you will most likely have consistent work for many months, or you could be put on small projects with temporary and limited work. As long as you deliver consistently good work you will have more opportunities.

It should be noted, that you could actually earn a living with a good project, or you could have spotty and incidental work. It’s not advisable to rely solely on one AI Platform for all of your income. It’s best to have other means of income to fall back on if the platform you’re working for, doesn’t have a project that is a good match for your skills.