Artificial Intelligence - Update with ChatGPT 04 14 2023
april 14, 2023 – Michael Marsh
We have been playing around with the ChatGPT engine.
The UI has been glitchy, likely a result of server overload. When the UI is able to send data to the AI engine, the responses are quick (1 - 5 seconds). My ratio when using the interface is running 1 success to 9 failures. This will of course be remedied as volume is baked into the cake and as free market providers customize the UI and engine to commercial customers.
Through the workers' compensation lense, the initial results showed that the training and data not yet ready for common use in the workers' compensation claims area.
There appears to be ongoing data and learning however. Below is an interaction with the ChatGPT AI engine about ways to decrease the duration of workers' compensation lost time claims.
"As an AI language model, I can provide you with ten general ways to decrease the number of days for a recovering worker in the workers compensation system to return to their time of injury job:
- Early Intervention: Ensure that injured workers receive prompt medical attention and intervention as soon as possible. Early treatment and management of injuries can help prevent further complications and speed up the recovery process.
- Effective Communication: Ensure that there is clear communication between the injured worker, their employer, the workers' compensation insurance company, and the healthcare providers involved in their treatment. This can help ensure that everyone is on the same page regarding the worker's recovery and return-to-work plan.
- Return-to-Work Programs: Implement a structured and supportive return-to-work program that provides injured workers with a modified or alternative work schedule or duties, if needed. This can help workers gradually ease back into their pre-injury work routine and avoid prolonged absences from work.
- Vocational Rehabilitation: Offer vocational rehabilitation services to help injured workers develop new skills or find suitable work if they are unable to return to their pre-injury job.
- Workplace Safety: Ensure that the workplace is safe and free of hazards that could cause injuries or re-injury to workers. This includes providing proper training and equipment to workers to help prevent injuries.
- Regular Check-ins: Maintain regular contact with injured workers to monitor their progress and provide support as needed. This can help workers feel valued and supported and may motivate them to return to work sooner.
- Flexibility: Be flexible in accommodating the needs of injured workers, such as offering flexible work hours or allowing for work-from-home arrangements if possible. This can help workers maintain their work-life balance and reduce stress, which can aid in their recovery.
- Case Management: Employ a case manager or coordinator to oversee the worker's compensation process and ensure that all parties involved are following the appropriate procedures and protocols.
- Empathy and Understanding: Show empathy and understanding towards injured workers and their families, and provide them with the necessary resources and support to aid in their recovery. This can help workers feel valued and supported, which can aid in their recovery.
- Review and Evaluation: Regularly review and evaluate the workers' compensation process and return-to-work programs to identify areas for improvement and implement changes that can help injured workers return to work more quickly and efficiently."
The rudimentary concepts returned by the AI are at least consistent with my non-machine learning over a number of decades. The improvement and accuracy are hugely improved over just a few months ago. AI may well have a place...but for now...workers' compensation remains a human helping human business.
Takeaway:
AI is not ready for primetime in the workers' compensation claims space. But improvements are noticed with increasing frequency and decreased cycle times between improvements.
However, I remain steadfast in my current communication with clients considering for software licensing in the AI space: Buyer Beware