Government AI 'Gennai' Deployed to Disaster-Stricken Areas: Is It the AI or the Practical Application That Saves Disaster Response?

Government AI 'Gennai' Deployed to Disaster-Stricken Areas: Is It the AI or the Practical Application That Saves Disaster Response?

When a large-scale disaster occurs, disaster-stricken areas face not only visible damage but also a 'flood of information'.

The situation at evacuation centers.

Road accessibility.

Power outages, water supply interruptions, and communication failures.

Shortages of supplies.

Information on the injured and those requiring assistance.

Notifications from national and local governments.

Inquiries from residents.

Field staff, who may themselves be victims, must collect, organize, and communicate vast amounts of information to relevant agencies.

In this context, the Digital Agency has urgently provided 'Government AI Gennai,' a generative AI for government employees, to local governments and disaster response agencies affected by the Reiwa 8 Kumamoto Earthquake.

The target includes not only the affected local governments but also supporting municipalities, fire departments, medical teams, and volunteer organizations. The goal is to support the aggregation, organization, and sharing of disaster information through tasks such as drafting notifications, transcribing audio, and converting data formats. (Digital Agency)

I believe this is a symbolic move indicating that the use of generative AI in administration has begun to advance from a 'tool to create documents a little faster' to 'practical infrastructure that supports disaster response'.

What is most lacking at disaster sites is not information

During a disaster, a lot of information is gathered.

However, what is truly in short supply is not the information itself.

It is the people who organize the information.

The people who verify the content.

The people who deliver it to the appropriate recipients.

And the time to process it into a form that allows for decision-making.

For example, compiling reports from multiple evacuation centers into a single table.

Transcribing meeting audio and extracting decisions.

Rewriting notifications full of technical jargon into easy-to-understand language for residents.

Standardizing data formats that differ by municipality.

Each of these is a mundane task.

However, during a disaster, these 'mundane tasks' relentlessly drain the time and energy of staff.

This is exactly where generative AI shows its true worth.

Rather than having AI make the final decisions for disaster response, it prepares the materials for humans to make decisions in a fast, readable, and easy-to-handle format.

Used in this way, significant effects can be expected in reducing the burden on the front lines.

What 'Gennai' can do is not flashy. But it is important.

When people hear 'AI,' they might imagine future predictions or high-level decision-making.

However, what is expected this time is more realistic work.

Summarizing long documents.

Transcribing meeting content.

Creating drafts of reports.

Converting table and data formats.

Creating guidance documents for residents.

Removing duplicates from multiple pieces of information.

To be honest, it lacks the flashiness that looks good on the news.

However, from the perspective of someone who knows the field of IT operations, it is this kind of mundane processing that truly supports systems.

If people making critical decisions spend hours copying and pasting or formatting tables, they cannot do their actual jobs.

What should be left to AI is not human responsibility, but the tasks that steal human time.

However, 'distributing AI' is not a solution

There is something we must be careful about here.

Generative AI is not useful just by being provided.

At disaster sites, communication environments can become unstable.

There may be a shortage of computers and charging equipment.

Not all staff are accustomed to operating AI.

Rules are also needed regarding what information can be entered and how to handle personal or confidential information.

Furthermore, the text generated by AI may contain errors.

Names of evacuation centers.

Addresses.

Quantities of relief supplies.

Water supply times.

Road traffic information.

If AI makes even a one-character mistake in such information, it could affect the safety of residents.

That is precisely why final confirmation must always be done by a human.

AI is not the person in charge of the disaster response headquarters.

Even if it can become an excellent assistant, it cannot take on the final judgment.

The biggest challenge is preparation during normal times

Learning how to use a new tool only after a disaster has occurred.

This is a very difficult situation for the front lines.

If we are to truly utilize 'Gennai' for disaster response, we need to train during normal times.

For example, the following templates should be prepared in advance.

Instructions for summarizing evacuation center reports.

Instructions for creating notifications for residents.

Instructions for separating decisions from pending items in meeting minutes.

Instructions for converting multiple damage reports into a list.

Rules for entering data while excluding personal information.

Checklists for verifying AI output.

What is important in disaster response is not just the performance of the tool.

Who will use it?

What will it be used for?

How much will be entrusted to it?

Who will verify it?

If there is an error, how will it be corrected?

Only when these are decided does it become 'operation'.

A common issue in Japan's administrative DX is the feeling that the work is finished the moment a system is introduced.

However, from the perspective of the front lines, introduction is just the starting point.

I want AI to be used not to reduce staff, but to protect them

Municipal staff during a disaster face an extremely heavy burden.

Long working hours.

Dealing with residents.

Successive inquiries.

Reporting to higher-level agencies.

Confirming ever-changing situations.

And the anxiety that their own family and home might also be affected by the disaster.

The introduction of generative AI should not be reduced to mere talk of personnel reduction or efficiency.

If work time is shortened by AI, that time should be used for rest.

Used for time to listen to residents.

Turned into time to find people who have not received support.

AI should be used to reduce the burden on humans, not as a tool to make humans work even more.

If you get that wrong, the technology you went to the trouble of implementing will become a tool that drives the front lines into a corner.

Digital disaster prevention has entered a new stage

The emergency provision of 'Gennai' this time is an important step toward Government AI being used in actual disaster response.

The Digital Agency set up a disaster response headquarters on the day of the earthquake and has also confirmed the operational status of government-wide business environments and relevant systems. It is clear that they are trying to support disaster response from a digital perspective by combining the maintenance of such foundations with AI-based operational support. (Digital Agency)

In disaster prevention to date, electricity, water, food, roads, and communications have been considered important infrastructure.

From now on, 'information processing capability' will be added to that.

No matter how much information is gathered, it cannot be used if it cannot be organized.

No matter how excellent a support system is, it is meaningless if it does not reach the people who need it.

AI cannot remove rubble.

It cannot carry water either.

However, it can organize who needs what and where, and deliver it quickly to those providing support.

That reduction of a few minutes or tens of minutes could potentially save someone's life or livelihood.

Conclusion

I consider the emergency provision of Government AI 'Gennai' to be a positive initiative.

However, what is important is not the fact that 'AI was introduced'.

Did it really make things easier for the staff in the disaster-stricken area?

Did the provision of information to residents become faster?

Did the necessary information and supplies reach the people who needed support?

That is what should be evaluated.

The protagonist in disaster response is not AI.

It is the people who support the residents on the front lines.

I want AI to be a presence that quietly supports them so that they can concentrate on the work they are supposed to be doing.

I hope that 'Gennai' will not end up as mere buzz, but will become a mechanism that truly reduces even one burden for those working and living in disaster-stricken areas.

Digital disaster prevention is not about showing off the latest technology.

It is about delivering necessary information to necessary people, even in the midst of chaos.

And it is about not leaving those who are doing their best on the front lines isolated.