AI & Machine Learning Pain Points

Data Overview

1. 1

   Aggressive Censorship and Safety Guardrails

   Platform Policy×7

   “I think because in my prompt, I mentioned the word "children" so ChatGPT just censored and locked itself out mid-way through the generation”

2. 2

   Restrictive Usage Limits

   Pricing & Billing×6

   “I made the horrible mistake of actually using it. I gave Sol Ultra exactly two real tasks... That single task completely drained my usage limit”

3. 3

   Poor UX and Product Structure Confusion

   UX / Design×6

   “the increasingly confusing product structure, terminology, and user experience across the ChatGPT website, cloud Work system, new GPT desktop app, Scheduled Tasks, Projects, plugins, Sites, and related features.”

4. 4

   Data Privacy and Litigation Disclosure

   Privacy & Security×6

   “regular users’ ChatGPT conversations were ordered disclosed to the plaintiffs for searching and perhaps other litigation-related uses.”

5. 5

   Production Reliability and Logic Failure

   Reliability Issues×5

   “The recurring problems include losing previously established facts, inventing details, overlooking evidence already provided, giving commands incompatible with the stated...”

6. 6

   Inconsistent Media Output/Artifacts

   Output Quality×5

   “Am I telling you not to stare too hard at the flaws because it’s just previs? Absolutely.”

7. 7

   Chat Organization and Search

   UX / Design×4

   “Finding an old conversation is harder than having the conversation in the first place.”

8. 8

   Unfair Cache Write Pricing

   Pricing & Billing×4

   “For workloads like these, a cache write has absolutely no value... why am I paying extra to create a cache entry we'll never read?”

9. 9

   Poor Customer Support Responsiveness

   Support Issues×3

   “I realized the case had been closed on the fourteenth night after I began waiting for a reply. There had been no notification.”

10. 10

    Model Opacity and Judgment Drift

    Model Quality×3

    “staying outside a lab means your judgment drifts because the models are opaque.”

11. 11

    NSFW and Inappropriate Hallucinations

    Reliability Issues×3

    “NotebookLM generated a story about an assault when answering my prompt”

12. 12

    Prompt Fatigue

    UX / Design×3

    “I have come to find the task of writing prompts, critiquing the output for accuracy, and building with them to be mentally draining.”

13. 13

    Multi-vendor Management Overload

    Developer Experience×2

    “managing separate keys, SDKs, and billing across OpenAI plus a couple other vendors is already a headache.”

14. 14

    Training Infrastructure Bottlenecks

    Developer Experience×2

    “gpu utilization was sitting around 30% the entire run and epoch times were way slower than i expected... the loader was the bottleneck”

15. 15

    Agent Sequence Logic Testing

    Developer Experience×2

    “production agents fail on *sequences* a malformed tool call at step 7, a “done” that wasn’t done, a loop.”

16. 16

    Recommendation and Search Bias

    Output Quality×2

    “whenever I ask ChatGPT for suggestions on B2B software, it often brings up my competitors, but not my own company’s offerings.”

17. 17

    Model Instruction Drift

    Reliability Issues×1

    “Drifts away from established formats during long chats.”

18. 18

    Faulty Auto-moderation

    Platform Policy×1

    “My last story got auto-removed by IG’s detection system, and now I’m in “Instagram jail””

19. 19

    Enterprise Adoption Plateau

    Industry Trends×1

    “Enterprise AI has hit a plateau, and it's worth being honest about what that plateau actually is.”

20. 20

    Privacy and Staff Data Access

    Privacy & Security×1

    “it fully admitted that staff at openai were free to review my financial data.”