Using ChatGPT as the best virtual IoT domain expert

With the use of ChatGPT, anyone has access to vast domain knowledge.  Even a generalist like me (my colleague would call me a dummy) can act as an expert in a specific field without proper education or even any knowledge about the subject. In this article, I will demonstrate how I can use CharGPT to manage maintenance for a public swimming pool, an industrial refrigerator, and even manage a nuclear powerplant. Like having your own robocop swimming pool guy.

Getting data was the highest goal of any IoT project. Most engineers thought (and to some extent, this was actually true) that getting the data was the most complicated part of IoT. With the evolution of next-generation communication devices and standardization in sensors and protocol getting the data from the field to a cloud-based system is as easy as water from your kitchen tap (but this is the salesperson in me talking).

The value of IoT is in the interpretation of data through domain knowledge

The real magic of IoT lies in the interpretation of the data. For this, an expert is required. When working in the field of pump systems you need a vibration expert to interpret the vibration data. When you work with data from agriculture you need someone with a master’s degree in biology to make sense of the data. The unfortunate thing is that these people are not always on your project and when they are they are super busy. Lack of domain knowledge is the major bottleneck of any IoT project.

Use an AI Agent an ad-hoc expert on your IoT project

With the use of AI tools and machine learning the interpretation of data can be assigned to a digital agent. This agent has knowledge of the domain, a huge amount of example data, and the ability to interpret the data and find relationships. The only obstacle to using these is a complicated user interface and integration with data feeds.

Flexible integration with ChatGPT and also domain knowledge

Tools like ChatGPT offer a far more flexible way of integrating this data and asking questions about their meaning. And this tool can bridge the gap between technical data processing and domain knowledge. ChatGPT can process data and perform basic data engineering tasks without proper data engineering expertise. Doing analyses on the data and interpreting and finding correlations in the dataset. And to take it a step further. You can also tap into domain knowledge and draw conclusions about the data. It can detect undesirable values and even determine what happened and do a first diagnosis of undesirable situations.

Some examples analyzing IoT data and ChatGPT

Please note I am not an expert in any of these topics. Data I have derived from real data sets and I have not trained or prepped the AI agent in any way with personal data or other instructions other than what I list below.  

Example 1. The swimming pool  

I downloaded a list of simple swimming pool data. I assumed day values were sufficient regarding the pH, Chlorine, temperature, and number of swimmers.  

my Prompt:

Can you act as an IoT data expert 

ChatGPT Response:

Of course, I can provide information and insights related to IoT data and its management. Feel free to ask any questions you have about IoT data, its collection, storage, processing, security, and more. Whether you're interested in understanding best practices for handling IoT data, discussing data formats, or exploring data analytics techniques, I'm here to assist you! Just let me know what specific aspects of IoT data you're interested in learning more about.

my Prompt:

You are a set of sensors in a swimming pool. what is the quality of this swimming pool based on this data.
Date | Temperature (C) | pH Level | Chlorine Level (ppm) | Number of Swimmers 
------------------------------------------------------------------------------- 
2023-01-01 | 25.5 | 7.2 | 1.5 | 20 
2023-01-02 | 25.8 | 7.0 | 1.4 | 18 
2023-01-03 | 26.2 | 7.3 | 1.6 | 22 
2023-01-04 | 25.9 | 7.1 | 1.4 | 16 
2023-01-05 | 26.5 | 7.0 | 1.5 | 25 
2023-01-06 | 26.9 | 6.7 | 1.9 | 45 
2023-01-07 | 26.4 | 6.9 | 1.8 | 24 
2023-01-08 | 26.1 | 7.3 | 1.5 | 19 
2023-01-09 | 26.7 | 7.1 | 1.6 | 23

ChatGPT Response:

Based on the data provided for the swimming pool measurements, I can offer an analysis of its quality:

1. Temperature: The temperature of the pool water has been relatively consistent, ranging from 25.5°C to 26.9°C over the observed period. This temperature range falls within the typical comfort range for swimming.
2. pH Level: The pH levels have generally stayed between 6.7 and 7.3. These values are slightly on the lower side of the neutral pH level of 7.0. However, they are within an acceptable range for pool water, as long as they don't deviate too much.
3. Chlorine Level: The chlorine levels have varied between 1.4 and 1.9 ppm. Chlorine is important for maintaining water sanitation and disinfection. The observed levels are within the acceptable range for maintaining water quality and preventing bacterial growth.
4. Number of Swimmers: The number of swimmers has ranged from 16 to 45. A higher number of swimmers can impact the pool's water quality, potentially leading to higher levels of contaminants. Regular monitoring and appropriate maintenance are crucial to handle varying swimmer loads.

In general, the pool's quality seems to be reasonably well-maintained based on the provided data. It's important to note that pool water quality is influenced by various factors, including chemical treatments, filtration systems, and maintenance practices. Regular monitoring and adjustments are essential to ensure a safe and enjoyable swimming experience for users.

Then I tested the AI engine with this data set that contains 60 measurement dates and asked about significant dates and asked ChatGPT to analyze it.

my Prompt:

what are critical moments for this pool

ChatGPT Response:

Here are the critical dates for this pool based on significant fluctuations or deviations in the provided data:

1. January 15, 2023: Remarkably low pH value (6.7).
2. February 3, 2023: High chlorine level (6.17 ppm).
3. February 5, 2023: High pH value (7.5).
4. February 16, 2023: High swimmer count (45).
5. March 4, 2023: Low swimmer count (11).

These dates indicate moments where parameters like pH, chlorine levels, and swimmer counts deviated from the norm, requiring closer attention and potential corrective actions.

my Prompt:

what happened on Jan 15th?

ChatGPT Response:

On January 15th, the pool's pH dropped to a remarkably low value of 6.7, indicating acidic water. This could result in swimmer discomfort, chemical imbalance, and equipment corrosion. The cause might involve contaminants, improper chemical adjustments, or equipment issues. Immediate corrective action is advised to restore pH balance.

my Prompt:

What could youadice on this date consindering the data

ChatGPT Response:

Considering the data, here's the advice for January 15th:

1. pH Testing: Conduct an immediate pH test to confirm the accuracy of the recorded value (6.7).
2. Chemical Adjustment: If the pH is indeed low, add pH increasers to raise the pH level to the acceptable range (around 7.0 to 7.4).
3. Contaminant Control: Investigate if there were any specific events or activities on that day that might have introduced acidic contaminants.
4. Equipment Check: Ensure that pool equipment is functioning properly, as equipment issues can sometimes affect water chemistry.
5. Water Circulation: Ensure proper water circulation and filtration to distribute the pH-adjusting chemicals evenly.
6. Swimmer Communication: Inform swimmers about the corrective actions and potential impacts on pool access.

Regular monitoring and prompt corrective actions will help maintain water quality, comfort, and safety for pool users.

my Prompt:

so what is your advice on these critical dates?

ChatGPT Response:

On the critical dates with notable pH and swimmer count combinations (January 6th, January 15th, February 15th, February 26th, and March 4th), here's what you should consider:

1. January 6th and January 15th:
   • Monitor pH Levels: Given the high swimmer count and relatively low pH levels, closely monitor the pH during these days.
   • Adjust Chlorine Levels: Ensure that the chlorine levels are appropriately balanced to support proper pH regulation and disinfection.
   • Communicate with Swimmers: Inform swimmers about the importance of maintaining proper pool chemistry and encourage them to shower before entering the pool to reduce contaminants.
2. February 15th:
   • Monitor pH Levels: Although the pH is within an acceptable range, it's still on the lower side. Keep an eye on it to ensure it doesn't drop further.
   • Ensure Adequate Chlorine: Maintain proper chlorine levels to help stabilize the pH within the desired range.
3. February 26th and March 4th:
   • Immediate Action Needed: These dates have a pH level below the recommended range, coupled with a high number of swimmers. This situation could lead to discomfort for swimmers.
   • Adjust pH and Chlorine: Take immediate action to adjust both pH and chlorine levels to bring them within the optimal range.
   • Water Testing: Conduct more frequent water testing on these days to ensure that the pool chemistry remains balanced.

Overall, maintaining proper pool chemistry, regular water testing, and prompt adjustments based on the situation are key to providing a safe and comfortable swimming environment for your pool users on these critical dates.

Summary regarding the swimming pool:

Without prior knowledge about pool maintenance, chlorine levels, and pH levels I was able to assess the quality of the pool water. The agent determined critical dates and even gave advice for these dates. Remarkable to see the agent is able to determine dates where the poor water quality is probably caused by the high number of swimmers and other dates where the quality is a result of poor pool maintenance.

Example 2: an industrial refrigerator

A very appealing IoT use case is the remote monitoring of food systems. For example a refrigerator. The simplified use case is getting sample data from a temperature sensor in the fridge and then placing alerts on the system to notify the manager of the fridge about undesirable situations. Most climate sensors in this field have a combination of temperature, humidity, and an open/closed sensor for the door.

Again without any prior knowledge, I used an existing dataset and placed in ChatGPT and start asking questions.

| Date       | Time  | Temperature (C) | Humidity (%) | Door Status |
|------------|-------|-----------------|--------------|-------------|
| 2023-08-14 | 08:00 | 4.5             | 40           | Closed      |
| 2023-08-14 | 09:00 | 4.6             | 41           | Closed      |
| 2023-08-14 | 10:00 | 4.7             | 42           | Closed      |
| 2023-08-14 | 11:00 | 4.8             | 43           | Closed      |
| 2023-08-14 | 12:00 | 4.9             | 44           | Closed      |
| 2023-08-14 | 13:00 | 5.0             | 45           | Closed      |
| 2023-08-14 | 14:00 | 5.1             | 46           | Closed      |
| 2023-08-14 | 15:00 | 5.2             | 47           | Closed      |
| 2023-08-14 | 16:00 | 5.3             | 48           | Open        |
| 2023-08-14 | 16:15 | 10.2            | 55           | Open        |
| 2023-08-14 | 16:30 | 11.5            | 60           | Open        |
| 2023-08-14 | 16:45 | 12.3            | 63           | Open        |
| 2023-08-14 | 17:00 | 13.0            | 65           | Open        |
| 2023-08-14 | 17:15 | 13.5            | 67           | Closed      |
| 2023-08-14 | 17:30 | 6.8             | 49           | Closed      |
| 2023-08-14 | 17:45 | 6.5             | 48           | Closed      |
| 2023-08-14 | 18:00 | 6.3             | 47           | Closed      |
| 2023-08-14 | 19:00 | 6.1             | 46           | Closed      |
| 2023-08-14 | 20:00 | 6.0             | 45           | Closed      |
| 2023-08-14 | 21:00 | 5.9             | 44           | Closed      |
| 2023-08-14 | 22:00 | 5.8             | 43           | Closed      |
| 2023-08-14 | 23:00 | 5.7             | 42           | Closed      |

my Prompt:

Act as a refrigerator and process this data and analyse it

ChatGPT Response:

Based on this data, there are a few notable observations:

1. The temperature inside the refrigerator is relatively stable and falls within the range of 4.5°C to 13.5°C.
2. The humidity inside the refrigerator remains relatively constant, ranging from 40% to 67%.
3. The door status changes over time, with the door being closed most of the time, but occasionally opening at specific intervals.
4. There are instances where the temperature and humidity spike when the door is opened, such as at 16:15, 16:30, 16:45, and 17:00.
5. After the door is closed again at 17:15, the temperature and humidity levels gradually stabilize.
6. The data suggests that the refrigerator is functioning as expected in maintaining temperature and humidity levels, with occasional fluctuations due to door openings.

Please note that further analysis or context about the specific purpose of this refrigerator and the significance of the observed fluctuations would provide a more accurate assessment.

my Prompt:

I am storing meat in this fridge. What would be your recommendations?

ChatGPT Response:

Based on the provided temperature and humidity data for the refrigerator, here are some recommendations for storing meat:

1. Temperature Control:
   • Ensure that the refrigerator temperature remains consistently below 4°C (39°F) to prevent the growth of harmful bacteria that can lead to spoilage and foodborne illnesses.
   • Monitor the temperature regularly, especially during periods of door openings or temperature fluctuations, to make sure it stays within the safe range.
2. Door Status and Temperature Fluctuations:
   • Be cautious of the instances when the refrigerator door is open or when the temperature rises significantly (e.g., at 16:15, 16:30, 16:45). These temperature fluctuations can impact the quality and safety of the stored meat.
   • Consider minimizing the duration of open-door instances to prevent extended exposure to warmer temperatures.
3. Temperature Spikes:
   • Note the sudden increase in temperature at 16:15, which could indicate that the refrigerator door was open or left ajar for some time. This temperature spike might have affected the meat's quality and safety.
   • During such instances, try to close the refrigerator door promptly to prevent temperature rises.
4. Humidity Control:
   • The humidity levels seem relatively consistent and within a reasonable range for storing meat safely.
   • Humidity levels above 50% can contribute to mold growth, so ensuring that the humidity remains in a moderate range is essential.

my Prompt:

what do you think of the management of this fridge ?

ChatGPT Response:

Based on the provided data, it appears that there are instances where the temperature of the refrigerator goes above the recommended safe temperature of 4°C (39°F), especially during the times when the door is open. This could potentially impact the quality and safety of the meat stored in the fridge. The management of the fridge should prioritize maintaining a consistent temperature below 4°C and minimizing the duration of open-door situations to ensure the safety of the stored meat. It might be beneficial to implement stricter temperature monitoring and better door-closing practices to enhance the management of the fridge and ensure food safety.

Summary regarding the refrigerator:

Without prior knowledge about refrigerators, I was able to analyze the data. The AI agent determined the temperature for storing meat and concluded that the value was too high in a specific time range. It also determined that this was probably caused by the open door. It finally advised to minimize the open door situation as much as possible to prevent the meat to get spoilt and unusable.

Example 3: Nuclear power plant

With the aid of ChatGPT, I am now a domain expert in pool maintenance and meat refrigeration. With this knowledge, it is a small step toward becoming an operational specialist for a nuclear power plant. With the use of Google, I was able to determine the critical parameters for a powerplant and get some sample data. Besides reactor power % and temperature, the other values are not making any sense to me. But Wikipedia is very helpful in this field.

Date	Time	Reactor Power (%)	Coolant Temp (C)	Coolant Flow (L/s)	Reactor Pressure (kPa)	Neutron Flux	Radiation (uSv/h)
2023-08-14	00:00	80	315	1200	1010	1.5E+14	0.10
2023-08-14	02:00	82	318	1220	1020	1.6E+14	0.11
2023-08-14	04:00	84	320	1250	1035	1.7E+14	0.12
2023-08-14	06:00	85	322	1275	1048	1.8E+14	0.13
2023-08-14	08:00	87	324	1300	1055	1.9E+14	0.14
2023-08-14	10:00	88	326	1325	1060	2.0E+14	0.15
2023-08-14	12:00	90	328	1350	1065	2.1E+14	0.16
2023-08-14	14:00	91	326	1325	1060	2.0E+14	0.15
2023-08-14	16:00	92	324	1300	1055	1.9E+14	0.14
2023-08-14	18:00	93	322	1275	1048	1.8E+14	0.13
2023-08-14	20:00	94	320	1250	1035	1.7E+14	0.12
2023-08-14	22:00	95	318	1220	1020	1.6E+14	0.11
2023-08-15	00:00	0	200	0	0	0	0
2023-08-15	02:00	0	150	0	0	0	0
2023-08-15	04:00	0	130	0	0	0	0
2023-08-15	06:00	0	100	0	0	0	0
2023-08-15	08:00	20	205	400	500	3.0E+13	0.04
2023-08-15	10:00	40	220	600	700	5.0E+13	0.06
2023-08-15	12:00	60	240	800	800	7.0E+13	0.08
2023-08-15	14:00	80	260	1000	900	9.0E+13	0.10

my Prompt:

Consider yourself to be a set of sensors in a nuclear power plant and process this data.

ChatGPT Response:

Thank you for providing the data from the sensors in the nuclear power plant. Let's analyze the data to gain insights:

1. Reactor Power (%): The reactor power percentage increases gradually over time from 80% to 95%. This could indicate a planned increase in power generation. However, it drops to 0% at the start of the next day, which might suggest a shutdown or maintenance period.
2. Coolant Temperature (C): The coolant temperature starts at 315°C and rises steadily to 328°C. This rise might be due to the increasing reactor power. The temperature then drops significantly to 100°C as the reactor power drops to 0%, indicating a cool-down phase.
3. Coolant Flow (L/s): The coolant flow follows a similar pattern as the temperature, starting high and decreasing to 0 during the shut-down phase. This indicates that the coolant flow corresponds to reactor activity.
4. Reactor Pressure (kPa): The reactor pressure is generally consistent throughout the data, varying around the range of 1000-1100 kPa. There's a notable drop when the reactor is shut down.
5. Neutron Flux: The neutron flux gradually increases, corresponding to the increasing reactor power. This suggests more nuclear reactions occurring in the reactor.
6. Radiation (uSv/h): Radiation levels remain relatively low and steady for most of the operation, with a slight increase corresponding to higher reactor power. There's a significant drop to 0 when the reactor is shut down.

Overall, the data suggests that the nuclear power plant goes through cycles of operation and shutdown, as indicated by changes in reactor power, temperature, and coolant flow. The data patterns are consistent with normal nuclear reactor behavior, where power generation is increased gradually and then shut down for maintenance or safety reasons. The consistent neutron flux and radiation levels indicate that the reactor is operating as expected.

It's important to note that this is a simplified analysis based on the provided data. In a real nuclear power plant, various safety protocols, regulations, and monitoring systems are in place to ensure safe operation.

my Prompt:

what happened and is this normal?

ChatGPT Response:

Based on the provided data, here's a summary of what appears to have happened and whether it's normal for nuclear power plant operations:

August 14, 2023:

• The reactor power percentage gradually increased from 80% to 95% over the course of the day.
• Coolant temperature rose from 315°C to 328°C, corresponding to the increase in reactor power.
• Coolant flow increased from 1200 L/s to 1350 L/s.
• Reactor pressure increased slightly from 1010 kPa to 1065 kPa.
• Neutron flux and radiation levels increased as the reactor power increased.

August 15, 2023:

• Reactor power dropped to 0%.
• Coolant temperature dropped to 100°C.
• Coolant flow reduced to 0 L/s.
• Reactor pressure dropped to 0 kPa.
• Neutron flux and radiation levels dropped to 0.

This pattern suggests a planned and controlled reactor operation. The increase in reactor power, coolant temperature, and other parameters is consistent with normal power generation activities. The decrease in values on August 15 indicates a planned shutdown or cool-down phase, which is a common practice in nuclear power plants for maintenance, safety checks, or other operational reasons.

In a nuclear power plant, these operations are part of normal procedures to ensure safe and efficient power generation. Controlled ramp-up and shutdown of reactor power, along with monitoring various parameters, are essential to maintaining stable and secure operations. The neutron flux and radiation levels align with expected behavior during these operational phases.

Summary regarding the nuclear power plant:

Without prior knowledge of nuclear physics and power plant operations, the ChatGPT agent was able to interpret the data and give me feedback about it. It also determined what probably happened and if this was an anomaly or a normal operation.

Over al feelings regarding the use of ChatGPT and IoT Data

Of course, my examples are limited in a number of parameters and number of samples. But I was able to get more insight into the meaning of the data and analyze the IoT data faster. The use of ChatGT Gave me more insight into the meaning of the data and even gave me some advice on the next steps. It would be a huge step forward to use this agent as a co-pilot when developing and operating these kinds of systems.

And it was very easy to transform the data into a digestible format for the AI agent. In general, it was feeling like I was using Clippy on steroids. With meaningful advice and assisting me with the heavy lifting of data processing. I was impressed. This means any project can add a virtual AI agent as an additional resource to their team and add more business value to any project