Today's Senate Hearing On AI Regulation & Impact On Open Source Models

Today’s Senate Hearing On AI Regulation

I tuned in a little late for the hearing but many interesting conversations going on between the committee and Sam Altman, Professor Gary Marcus, and Christine Montgomery from IBM.

One of the most interesting things I found was that of the three, the one that does not want any new regulations is IBM. Early on in a battle with Lyndsey Graham, Ms. Montgomery tried to explain her position but Graham kept interrupting and talking over her.

Frequently, the current AI technology was compared to the car, the printing press, and the atom bomb. All of which had very different impacts on society.

The overall consensus from everyone is the cat is out of the bag with AI and we have no chance of getting it back. I totally agree, in the short time it has been in public hands t has been transformative. Transformative in both good and bad ways.

While Sam Altman is in favor of regulation, he also believes that society is able to identify “fakes” or ingenuine content over time. The question is whether the technology is developing faster than our ability to identify the fakes. He compared it to photoshopped photos and how after a while we were able to identify if something seemed photoshopped.

I think the other side of that is we also have grown to accept the fact that a photo has probably been photoshopped. I think the biggest problem is while many people will be able to identify ‘fake’ or misleading content, a majority of people will believe what they are reading or seeing is real. This content already spans text, video, audio, and images and can proliferate at a rapid rate.

The Coming Election Year

Everyone on the committee is very concerned about the impact of the coming election and the possibility of misinformation on a scale previously unimaginable.

Senator Klobuchar gave an example of a prompt she gave ChatGPT to write a tweet about a polling house being overwhelmed with long lines and where people should go instead. Her point of that was ChatGPT made up an address. In reality that was a bad prompt. As part of that prompt, the address of where they should go instead should have been included.

Most people on the committee have no idea how to effectively use the platform but understand the power and impact the platform can have.

Effective regulation will most likely not be in place before the next election cycle begins. We have already seen our first attack ad completely generated by AI, I expect many more will be coming from both parties.

Impact Of Regulation On Open Source Models

Amongst the talk during the hearing was an idea of licensing by companies using AI technology, as well as oversight and review before a product goes into production.

That raises a number of possible concerns:

• Will smaller companies be able to keep up with the regulation and oversight

• Will the larger players like OpenAI, Google, Meta, and Microsoft control the market and prevent further iteration by smaller companies?

• If we don’t establish a national regulation, regulation on the state level (already started) can make the ability to be compliant even more difficult.

• How can we create an environment with protection in place while still competing in a global market?

Open-source models are currently keeping OpenAI and Google moving forward quickly. The Information reported that OpenAI is now planning to release an open-source model and that may force Google to do the same in the near term.

Recording Of Live Broadcast On CNBC

I am in the process of summarizing the discussion and compiling key takeaways I’ll share tomorrow.

Until tomorrow,
Kevin Davis
760-835-8347