Did you that the retail company Gap is expecting net sales growth in the coming year? I heard about during that new Microsoft video where they demoed ChatGPT! It looks really great until you realize there’s no way clothing retail is going to go up at all in our suffering economy and ChatGPT lied to me!

And Google Bard? More like Google Bad! But it’s not just Google guys! It’s the AI craze everyone has been talking about and the chat bot to triumph over all chat bots, ChatGPT. AI has a place in automating grunt work that could otherwise be incredibly annoying to do. I use a few offline programs from such purposes, but what ChatGPT does completely breaks the purpose of AI and it’s time to expose its weaknesses and why I have not and will never touch it for the remainder of my life!

ClosedAI

First, let’s talk about OpenAI, the company responsible for the creation of ChatGPT. OpenAI was originally a non-profit organization that wanted to collaborate with other businesses and groups to expand the use of AI, but to keep things short, OpenAI spent hundreds of millions of dollars to build their organization’s infrastructure and were unable to make any money and as a result, became a “capped for-profit” company.

We Don’t Know What’s Happening…

Lots of big names have invested in OpenAI for years, but most importantly are the products developed by OpenAI, like:

  • Alignment: an AI traning framework
  • DALL-E 2: an image generation program
  • Whisper: a transcription tool
  • And ChatGPT: the informational chatbot.

The problem is out of every tool released by OpenAI, we know little to nothing about how any of their programs work. In fact, out of every program I mentioned, Whisper is the only one that is fully open-source. In fact, Whisper became so popular they removed the issue tracker from their GitHub. So much for feedback, right?

OpenAI provides technical documentation and research papers for their tools, so we know roughly how these programs work, but we don’t know the process by which OpenAI trains anything. It’s really on the basis of “just trust us”, but that’s not good enough in my book.

The Time Exclusive

In fact, some Time reporters thought they got the 1-Up on OpenAI by describing how OpenAI hires Kenyan workers to filter out toxic waste from their training models for less than $2 USD/hour.

First off, us brain-dead Americans and Westerners expect to get more money, but this amount that they are paying is actually quite fair and it’s much higher than the standard wages of Kenya. The issue is, if Time can go run with this dumb story like it’s some sort of secret, it really shows how little anyone outside of OpenAI really knows what’s underneath the surface.

This isn’t even touching on the “open-source” thing, because why would they open-source anything? But it’s more about how is it that a company called OpenAI never talks about publicly or lets anyone see behind the curtain.

Open In the Public

But let’s say we don’t care about the nitty-gritty technical details—what about their privacy? I don’t think it’s fair to test OpenAI’s security since they haven’t been open to the public for a long time, but privacy I believe is something they can fairly be judged for.

Please note that phone verification is required for signup. Your number will only be used to verify your identity for security purposes.

A serious crime OpenAI commits is demanding for your phone number or as we should take to calling it, your internet social security number, because every dang service wants your phone number, the personal tracking beacon of your life, to do the most basic things online! It’s already bad enough in the “land of the free and home of the brave,” but other countries have it worse. You basically need to hand over your real identity to OpenAI.

I refuse to use any service that demands a phone number out of me and I can’t wait for someone to say in the comments “Your number will only be used to verify your identity for security purposes.” The bleak reality is OpenAI couldn’t give one about your privacy. In fact, reading through their privacy policy yields legal boilerplate that lawyers love, great things like:

Conducting/Sharing Research About You:

To conduct research, which may remain internal or may be shared with third parties, published or made generally available;

Using PII to Feed the Beast:

To develop new programs and services;

Sharing Personal information

In certain circumstances we may share your Personal Information with third parties without further notice to you, unless required by the law, including without limitation in the situations below:

  • Vendors/Sevice providers
  • Business acquisitions
  • Affiliates
  • Other Users

But none of these things could really be that bad, right? Who could OpenAI possibly affiliate with?

Microsoft Marketshare

Microsoft has been very open about their integration into OpenAI with executives proudly bragging about the integration ChatGPT will have in Bing and AI throughout the rest of Microsoft AI products.

But we all know Microsoft is a soulless corporation that wants in on the next best thing and has transformed the tech industry into a mad rush to get AI in an arms race with Google and all of us civilians are caught in the crossfire. Microsoft isn’t content on just integrating in Bing, they want to control OpenAI itself and harvest the wealth of data and information of its users. Microsoft has invested $10B into OpenAI, outpacing all of the other donors of OpenAI and owning the absolute maximum of 49% of the company.

Microsoft’s Acquisition Addiction

It isn’t the AI or the open one they dumped money into; it’s about setting the stage for market dominance. In the past year, Microsoft has been ravenously hungry to capitalize on their dominance outside of their already thoroughly crushed enterprise field.

  • One of the worst news of last year was the $69B acquisition of the popular games company, Activision/Blizzard and $7.5B Zenimax, aggressively trying to centralize a grip on gaming communities.
  • Microsoft has been expanding its advertising network, specifically Bing’s own advertising network, by signing a major deal as the principal advertising of Netflix’s new free tier.
  • Even though most people previously laughed at Bing, Bing was still consistently the second most popular search engine to Google. While Google is imperfect, at least Google isn’t objectively evil like Microsoft is.

And all of these factors positions Microsoft to extend their reach to control more of people’s lives. And what better time when all of other Big Tech companies are in the federal wheelhouse to regain the power they once had…

ChatGuineaPigTesting

An unfortunate side effect of everyone talking about how amazing ChatGPT, and this isn’t to downplay it, sometimes it is awesome. But when their system can get so many things wrong, even under the prerecorded fake demos created by Google and Microsoft, can we even trust this technology?

ChatGPT and GPT-3 was and has been trained at this for a long time and pushed it out to the public as some sort of guinea pig test, and it still gets everything wrong.

On Wikipedia, there’s a screencap where someone asked ChatGPT if Jimmy Wales died in the Tienanmen Square massacre, which it correctly answers no, but it wrongly states Jimmy Wales was 23 at the time. The core question is answered, but accuracy is an ongoing issue with many prompts you issue ChatGPT.

Here’s some more good ones:

It gets better, it adds this cute little disclaimer at the bottom, claiming “oh it’s not a definitive answer” (and if it’s the Bing one, it’ll probably add an emoji afterwards), but this is a weak excuse for leading people to overvalue the system and treat it as an authoritative source,

because they just want their answer.

Sometimes, the AI gets derailed and seemingly for no reason.

Then there’s also the problem of legal issues and the fact that Bing giving this training network the internet, the AI is now going to be trained on material people never would have imagined or consented to.

For example, ChatGPT and GitHub Copilot can generate code if you ask it, but what if this code was originally from somewhere and sometimes violates the license of the original code? If you ask ChatGPT to write you a story, how do you know that ChatGPT isn’t pulling from material that could hold you liable for plagiarism of intellectual property? Guess you better have fun getting sued to find out!

Then there’s all those Ne’er-do-wells using it in malware or trying to exploit governmental systems. In fact, this has gotten so bad that Microsoft has had to limit questions to 5 heavily neutered responses and sneaking in emojis at the end. That’s going to get those pesky troublemakers to stop! I’m sure the consequences were considered!

Outro

Okay, I’m done. I’m never touching any of this Bing, Google Bad, OpenAI crap for a long time.

Additional Videos: