This web app uses cookies to compile statistic information of our users visits. By continuing to browse the site you are agreeing to our use of cookies. If you wish you may change your preference or read about cookies

April 21, 2023, vizologi

Open letter for model shutdown beyond GPT-4.

It seems an exercise in unmitigated hypocrisy on the Elon Musk-led hiatus or truce to stop LLM model training beyond GPT-4. First of all, Musk was one of the first investors to create OpenAI from scratch, and he made a lot of money on the exit deal when Microsoft entered the second round.

Secondly, weeks before the release of the letter, he created a new limited partnership called, made a massive booking to buy machines from Nvidia, and made an enormous call to recruit competing engineers to develop a new model that would compete directly with GPTs, three weeks later, he announces that we need to stop everything to prevent advanced AI systems from becoming risks to society and humanity.

Besides pointing the finger, at a specific company in the market and at a particular GPT-4 product version, the graphic below shows the ecosystem of existing models, maybe they are not as mediatic as ChatGPT, but in this scenario, there are several companies behind, and many public and private commercial interests, perhaps it would have been more reasonable to launch a more generic stop, for example, comment that it stops at models beyond a trillion parameters, you make it generic and do not target anyone, nor try to modify according to your interests the natural rhythm of the market and the state of the art of the technology.

LLMs Landscape

Then it is simply a business strategy in a market that has just been born, stop, that does not give us time for others to keep up; once the halt is lifted, we will bring out all our artillery. The Chinese will be rubbing their hands together, and I assure you that they will not stop their Wu Dao 2 developments.

Far from it, I am not the number one defender of OpenAI; my current technology partner and algorithm provider, which started as a non-profit open source company, duped the whole community, telling us a terrific movie back in 2016, and has been transformed over the years, into something that I do not even know what it is today, it takes an MBA to understand the current composition of the limited partnership of that company, which retains a 2% non-profit.

OpenAI deserves a separate article to understand what they are doing, and I know to some extent, that these corporate transformations they have been making, as well as the change in the company’s mission, may have been the trigger that has bothered Musk so much to start this initiative.

I do not see a genuine intention to tackle the problem, and I do not see an honest intention to protect the future of societies in this operation. But instead, a technical stop to redistribute the pieces on the board.

Is it necessary to open the debate now? Yes, of course, among all of us, users, experts, companies, entrepreneurs, legislators, and organizations, we should be thinking about impacts, benefits, risks, ethics, values, framework agreements, and how to manage this disruptive typhoon of generative AI that started two years ago and is already impacting the whole of society.

From calm, willing to agree, open conversation, and without crossing commercial interests, not from alarm, fear, control, and misinformation, a breeding ground for policies such as in Italy, which from ignorance and misinformation, prohibit the use of this technology, condemning its citizens to ostracism, this type of actions harm the entire ecosystem, and I insist that they are based on an unjustified alarm.

I know that regulation is needed to build fair and consensual rules that benefit all agents in the ecosystem, ensuring that we work safely. However, I do not want us to panic, be in a hurry, and start over-regulating something that we still do not understand very well how it works.

So far, I have expressed my opinion as Pedro, a citizen, technologist, and blogger. From here, I change my hat and give you my opinion as Pedro, founder of the first Spanish generative AI company. Here I tell you that I breathe a sigh of relief for the requested stop; I have been telling you since last year that the speed we were going was not expected, let them stop it for six months or a year, whatever it takes, but the pace was frenetic.

Managing a product with so many changes and tests was becoming hell; if we run so fast, something can break along the way, and here I speak as an engineer, who has already participated in a few technology projects, and I have seen almost everything.

I trust the testing, security, and filtering that OpenAI teams do before releasing a new release. I trust the testing that I do to ensure that the product we offer to our customers meets the necessary security standards.

But we started in January with updates to text-davinci-003, in March, without respite, we launched gpt-3.5-turbo, just after gpt-3.5-turbo-0301, and look we were already waiting for gpt-3.5-turbo-0302-GTI-16valve-ultra-boost, just before reaching the launch of GPT-4, which was two weeks later, sorry to express this information in technical terminology; those are the technical names below the trade names we all know, as ChatGPT.

My CTO was already preparing the 7 Shiva elephants to send them here to the office in Alicante to visit me. They are complex months to manage a startup in this scenario of continuous change.

The stop, if true, will come as water of May, to take a breath, and to be able to raise product roadmaps with sense a month ahead, no more, I only ask for a month, without changing anything, people, the end user also needs a break, they can not assimilate so much technology and exponentiality at once, and in the long term will be counterproductive for everyone, generating a dysfunctional blockade.

Then, Sam Altman (CEO of OpenAI), you are the best in the world in AI, the smartest, have already surpassed Google, and have practically taken 80% of the generative AI market.

You have shown us all that you are competent and brilliant, but hold on a little bit, and no one can stand it, neither the startups, nor the users, nor the society as a whole, let’s take advantage of the break, to reflect, and think about where we come from and where we are going.

It would be a good time for you to review the architecture of the systems. For example, suppose your system crashes in the first month due to insufficient planning. Although, in that case, it is understandable, according to the 100 million that came in all at once, if it crashes as a tradition six months later, it is in bad taste with the 11 billion of capitalization and the almost unlimited access to the cloud of Microsoft Azure.

Review the business model of ChatGPT, the first time in the recent history of innovation and startups which is segmented by a freemium denial of service; I still do not give credit, which should be a shame, have become a business model; I offer you a free service, I charge you for the premium so that the service does not fall, and it keeps falling both the free and premium. Moreover, I have everything in the same pull servers where I serve my partners and startups, and the error occasionally spreads and spreads to all.

Please implement SLAs (Service Level Agreements) for service interruption; you are not a startup with 11 billion of capital and almost 800 employees; no matter how much you say you are a startup, you are a big company that can not work at this level, and gives this image, without having service agreements when your systems are continuously down, it is a super primary issue, no matter how much you tell customer service, that they put the smile emoticon when they are asked about SLAs by customers, after continuous interruptions of services in the API.

And finally, it is also an excellent time to review the company’s communication; the misinformation about your company and products is brutal, and continuous leaks and hoaxes make it impossible for your customers who pay you to know where we are precise; it is an exercise in archeology. Then, apart from training the next GPT-5, there is a lot more work to be done in a company than just focusing on the latest version of GPT.

Vizologi is a revolutionary AI-generated business startegy tool that offers its users access to advanced features to create and refine start-up ideas quickly.
It generates limitless business ideas, gains insights on markets and competitors, and automates business plan creation.