Companies are training AI by tracking their employee’s actions and many don’t even know it. Advancements in AI training techniques allow computers to watch employees complete their jobs and simultaneously use the information to replace them.
It’s a questionable trend that is unlikely to go away, and brings up both economic and ethical considerations.
Artificial Intelligence models use a technique called Machine Learning to complete tasks. The computer essentially guesses and checks its answer countless times per second until it derives an answer. It’s a lot of complex math that happens incredibly fast. Even people that research AI and Machine Learning have difficulty explaining why an AI answers the way that it does.
For machine learning to work, computers need to be fed with extremely large amounts of data. Software like ChatGPT use text information from the internet. Next generation AI will use video recordings of employees completing tasks.
White Collar Workers are Already Training AI
If you work at a large company, chances are that the tasks you are completing on a daily basis are being tracked as part of an AI process mining initiative.
Software is either installed on user computers or is installed on a central server that tracks computer activity through logs, or more directly records interactions with popular software such as Salesforce and SAP.
The information is consolidated and used to identify repetitive tasks that employees are engaged in. The tasks are then categorized and broken into smaller pieces to identify if they are good targets for automation.
UiPath is a leader in robotic process automation, and is just one of the companies that provides AI process mining solutions to companies. The YouTube video below gives an overview of how this technology works in practice.
Solutions like UiPath represent a first-generation approach to AI process automation.
Second generation tools are coming and will be much more invasive and advanced.
Robots were recently trained to complete tasks by watching YouTube videos and in much the same way, automation bots can be setup to record videos of user computers to identify how tasks are being completed. The video recordings can be fed into a Machine Learning model to effectively replace human machine interactions.
Is it ethical to train artificial intelligence without employees knowing?
As more companies lean on their existing workforce for training data that feeds into AI models targeted at replacing them, we have to wonder if it’s ethical to do so without employee. knowledge.
Many companies will likely have legal cover with generic employee handbook clauses like the following:
“All employee communications, including but not limited to telephone conversations, email exchanges, and computer interactions, are subject to monitoring and recording. This is done to ensure adherence to our standards of service, quality control, and for training purposes. By accepting employment with our organization, you acknowledge and consent to this ongoing monitoring and recording of your work-related communications and activities.”
Even being within their legal rights to record and track employee digital communications, one has to wonder if employees would act differently if they knew that everything they did was to train an AI.
Some of the potential concerns we can see employees having include:
- Privacy Intrusion – Many employees may see it as an invasion of privacy to record every single action taken at work from checking the news or status of an Amazon order.
- Informed Consent – Employees may not know what they are consenting to until after they have accepted a job.
- Limitations of Purpose – Information used for AI training may exceed its initial scope and be used for other purposes that employees are not aware of.
- Misrepresentation – Data collected in the name of AI training, could be misinterpreted by an AI to flag an employee’s actions as harmful to the company resulting in unintended adverse actions or termination of emploment.
These are only a handful of the potential concerns that employers will need to address as they ramp up AI training programs. Even with the appropriate level of legal disclosures, employees may also feel a drop in morale and increased resentment knowing that they’re being asked to eliminate their own jobs.
Should Employees be Compensated for Training AI?
Employees who take on additional responsibilities and job duties often expect to be compensated for their work. Either directly or indirectly through recognition and increased opportunities for promotion.
As companies look for processes to automate and data to train them, they will find that not all employees are equal.
In the case of a call center or customer service agent, they may provide more satisfactory answers that should be used by a majority of other call center representatives.
The ability of people to provide better training data adds more value to a company, and ultimately should help them improve margins by eliminating and automating jobs.
AI Training will be a highly sought after skillset in the future.
Some employees will utilize computers and complete tasks in ways that are easier for a computer to interpret. Much like how schoolteachers have a specialized skill set to impart knowledge to children, an AI Trainer would have a specialized skill set to show computers how to best complete tasks and handle exceptions.
Conclusion
Companies are investing billions of dollars to implement AI in their organizations. Most initiatives are focused on cost savings, which ultimately means more automation and a less workers.
For AI to improve and meet corporate expectations, it will need to have access to large amounts of training data.
The data will be provided by existing employees who will largely be unaware of it.
Organizations should get ahead of it, by implementing updated employee handbooks and publish employee announcements so that employees are at least informed of the practice. At best, companies should compensate their employees that excel as corporate AI trainers for taking on the additional duties.
It will be a difficult line to walk as widespread AI automate initiatives have the potential to decimate employee morale as workers begin to see their ranks dwindle in the coming years.