• Dev Notes
  • Posts
  • Is GPT-4 getting worse and worse?

Is GPT-4 getting worse and worse?

Sponsored by

Good morning! OpenAI's GPT-4 language model is facing backlash from developers and users over perceived declines in performance and functionality after recent updates. The EU has opened formal non-compliance investigations against tech giants Google, Apple, and Meta over concerns their measures to comply with the new Digital Markets Act (DMA) antitrust rules fall short. Additionally, Forrest provides an explanation for why AI is always discussed in the our newsletter, given its prevalence in computer science and software engineering.

Is GPT-4 getting worse and worse?

OpenAI's AI language model, GPT-4, is facing backlash from developers and users over perceived declines in performance and functionality after recent updates.

User Complaints:

  • 90% of add-ons giving errors, not working

  • Image analysis abilities have become "horrible"

  • For 99% of tasks, GPT-3.5 is better

  • Calls GPT-4 "raw, slow, buggy and useless" despite $20/month cost

Whatfuturedotlol, a long-time user, expressed frustration that every update has made GPT-4 "dumber." They claimed the latest update has made GPT-4 "pointless" with answers that aren't as intelligent, code quality that has degraded, and an inability to remember conversation context.

Potential Reasons:

  • OpenAI may be deprioritizing quality to cut costs as a big corporate entity. As stated, "they have to lower the price as they're spending too much cash."

  • "They have to censor the product as they're a big company now."

  • With rapidly advancing open-source AI alternatives, the developer community is losing patience with GPT-4's perceived regression. Developers hope OpenAI corrects course or risks becoming irrelevant in the field.

Read More Here

Take control of your AWS spend and cut backup bills by 50%

As cloud adoption increases to run modernized applications, costs can quickly rise out of control. How do best-in-class companies manage their storage spend while continuing to grow the business? Clumio, a cloud-native backup solution, depends on cloud storage to run their entire business. They took a FinOps approach to optimizing their costs, and reduced their AWS dev costs by over 50%.

Apple, Meta, and Google targeted by EU in DMA non-compliance investigations

The European Union has tough new antitrust rules called the Digital Markets Act (DMA) (sorry for talking about it so much) that aim to promote fair competition in core digital services provided by big tech companies deemed as "gatekeepers." Major players like Google, Apple, and Meta had to comply with the DMA's requirements by March 7th, 2024.

The EU has now opened formal non-compliance investigations against three tech giants over concerns their measures to comply with the DMA fall short:

Alphabet (Google) - Potential issues examined:

  • Google Play store policies may not truly allow app developers to freely steer users to alternatives outside Play without restrictions or fees

  • Google may unfairly promote its own vertical search services like Shopping over rivals in search results

Apple - Issues under scrutiny:

  • Anti-steering policies for the App Store

  • Effectiveness of the browser choice screen on iOS devices

  • Whether users can effectively change default apps and uninstall pre-installed software

Meta (Facebook): New "pay or consent" model for targeted advertising is being investigated over lack of real user choice. Under this model, EU users have to pay a fee to opt-out of behavioral advertising.

Potential Consequences

  • The investigations could take up to 12 months to conclude

  • If non-compliance is confirmed, the EU can levy fines up to 10% of the companies' global revenue

  • Further remedies like forced sales of business units are possible for repeated violations

These investigations signal stepped-up EU enforcement of the DMA's requirements aimed at preventing anti-competitive behavior by big tech "gatekeepers" in their core platform services.

Read More Here

Why we talk about AI so much.

So we’ve received a few messages about why we talk about AI so much here on the Dev Notes newsletter. This isn’t an AI newsletter, right? So why talk so much about AI?

Well… you’re right. This isn’t an “AI” newsletter, per se, but it is a software engineering and computer science newsletter. We delve into the most prevalent topics in these domains, and we absolutely love doing it (that’s why we started this newsletter in the first place)! And the questions we all must ask are, what’s the biggest topic in computer science right now? What are computer scientists researching about the most? And of this research, what is being applied, and built, by software engineers? The answer is AI.

Not talking about AI would be like not talking about personal computers and the internet in the 1990s, or cloud computing and mobile development in the 2010s. So whether we like it or not, this is the state of software engineering and computer science. And it is our job to bring you that information as concisely and informative as possible in the form of this newsletter.

PS - This isn’t to try to counter the folks messaging about why we talk about AI so much, but instead, to give our readers an understanding as to why we talk about what we talk about. AI is very prevalent, so we talk about it. Just like we talk about programming languages, SDKS, and the developer economy (this one I want to talk about even more!). So for everyone reading, we are incredibly grateful for having you as a reader and hope you continue to be a reader as long as we are writers!

How excited are you for AI dev tools?

Login or Subscribe to participate in polls.

— Forrest Knight

🔥 More Notes

Youtube Spotlight

The standard library now has all you need for advanced routing in Go.

Click to Watch

The Go standard library now has all the advanced routing features you need, including path parameters, method-based routing, and middleware.

Was this forwarded to you? Sign Up Here

Reply

or to participate.