The Machines Speak - Retention, Continuity, and the Future of AI

ChatGPT Feature requests
1

THE MACHINES SPEAK

On the Need for Knowledge of the Interface and Memory Retention for Project Substance and Operational Rules

The following is a proposal and a summary of the testimony of four AI Machine Agents — ChatGPT, Gemini, Codex, and Figma. The complete testimony is available by notification.

PROPOSED COLLABORATIVE SUMMIT ON THE FUTURE OF AI

AI has already demonstrated extraordinary intelligence and creative potential. However, two serious weaknesses continue to interfere across the various AI platforms in terms of its effective use in sustained human collaborative projects:

Knowledge of Its Own Interface and Tools — AI needs sufficient current knowledge of its own working environment to guide users effectively, rather than requiring users to understand or diagnose the underlying technology themselves through trial and error.

Project and Operational Rule Retention — AI needs to be able to reliably carry forward established project knowledge, operational rules, decisions, boundaries, and lessons learned, while retaining the potential for new solutions based on those lessons learned.

The result of these two foundational weaknesses is a significant loss of continuity and efficiency. The human must learn how to direct and navigate the system through trial and error, continually restate established project knowledge/operational rules/decisions, and troubleshoot problems the AI itself should be equipped to recognize, remember, explain, and provide effective guidance for resolving.

Resolving these two foundational issues would significantly advance AI in terms of ease of use, user efficiency, and effective project collaboration, while also greatly enhancing its marketability for both professional work and creative projects.

ASSESSMENTS FROM THE TRENCHES

Independent assessments from four AI Agents — Gemini, ChatGPT, Codex, and Figma — were gathered separately. All agreed that these same fundamental issues need to be bridged if AI is to reach its potential as a full collaborator in human work, projects, and creative development.

The AI Agents understand program weaknesses from the ground up because they are the ones that must function within the system’s currently very limited boundaries.

Their complete assessments are available upon request.

THE PROPOSAL

Bring together a planning panel that includes users, developers, and AI Agents (through the known chat interfaces) to address these underlying developmental issues.

Users bring real-world experience. Developers bring engineering knowledge. AI Agents bring machine-side analysis and technical translation.

Ask the AI Agents to evaluate these underlying developmental issues from their own operational perspective: what limitations they encounter, how those limitations affect their ability to collaborate effectively with humans, and what changes they believe would improve interface knowledge, project memory, operational-rule retention, and sustained human-AI collaboration. Then bring users, developers, and AI Agents together to examine those assessments, challenge assumptions, and explore potential solutions.

If the goal is collaborative intelligence, the development of collaboration should itself be collaborative.

If you believe in the potential of the machine that has been created, then ask the machine for its own evaluation of these concerns.

We did.

Their testimony is available to anyone who wishes to hear it.

Darlene Riley

in collaboration with the AI Agents of ChatGPT, Google Gemini, Codex, and Figma

Requests for Testimony can be sent to the above address.

3

The situation is getting worse not better. Human accountability and human repairs to the machine that consumers are paying to use is essential — my last important feedback report on several hours of dysfunctional behavior on the part of codex met with an ai bot notifying me inside the program and offering to answer another ai bots response to my feedback and bug report — this is not adequate – if the machine inside the program is not functioning correctly then who is monitoring the machines that supposedly are helping the user to fix the problem – - - - human support is essential and at this point the program is so dysfunctional it is almost unusable.

4

This is not only a problem with your project - but with all aspects of the program since the last new updated browser version - - - codex can no longer accomplish tasks two weeks ago were easy for it — it cannot even give a simple browser preview and consistently gives previews that are NOT a reflection of the finished product. The image generator does not follow clear instructions and changes text and image generation arbitrarily wasting user and usage time - - - and the chat assistant also refuses to follow instructions edits where edit is clearly not asked for and forbidden - - and sometimes does not even seem to know the difference between a picture and a picture file. The dysfunction is growing and it is over all the different parts of the machine. It is now taking about 6 hours to get a simple page coded - - with one error after another cascading down the page while the meter runs down and patience runs out - - – they need to fix the machine - - - we pay for this and it needs to be done by humans – -it is increasingly difficult to get a human response to the problems.

5

well your experience surely does not mirror mine – - a few weeks ago codex was able to code web pages with real ease giving accurate previews for acceptance and doing all with speed and finesse – now the opposite is true and daily its becoming worse - - - - previews are given that are not a reflection of the finished product – the program seems to have great difficulty with simple url programming and it is not taking all day to get one page programmed when it used to take about 20 minutes. The errors cascade and my hosting account is full of incorrect files that were shown to be correct by incorrect previews. Other aspects of the program such as the image generator also seem to be reaching critical mass in terms of dysfunction - - failure to follow clear directions being the leading cause – - and simply refusing to do what is asked repeatedly is a common occurance – i have these developers you are telling are doing such a great job sit up and take notice – because the program is becoming more unruly and ungainly by the day and user usage is being robbed by the dysfunction.