The use of generative artificial intelligence services publicly available on the internet, such as ChatGPT , Copilot and Gemini, can pose risks to the security of companies due to possible leaks of information and work data.
A safer, and even economical, alternative would be to develop an AI internally, according to experts interviewed by FreeGameGuide at Dell Technologies World , the company’s annual event that began on Monday (20), in Las Vegas (USA).
Michael Dell , CEO and founder of Dell, is one of the advocates of “data sovereignty” and states that companies should have storage and processing infrastructures at least for the most sensitive data. Additionally, training an AI or Large Language Model (LLM) in-house can also be cheaper.
“Recent studies have shown that running generative AI on local servers can be up to 75% more economical than paying to do the same thing on public cloud services. That’s why 83% of companies surveyed plan to ‘repatriate’ some of their computing demands from the cloud to our own servers in 2024. This is happening for two reasons: inference and data gravity,” Dell told FreeGameGuide.
This year, Dell launched the IA Factory, a type of AI factory capable of training and running customized LLMs for any size company. That said, it is natural for the brand’s founder to defend the idea that its customers prefer their own AI produced using open source models and with infrastructure provided by Dell itself, rather than paying for a ready-made service from OpenAI or Google , for example.
“Data gravity” was also mentioned by Michael Dell. This concept indicates that data has the potential to attract and generate more data as decisions are made.
In this scenario, a company that pays for cloud storage is essentially trapped in an ecosystem that will charge more and more to keep its information safe. And if she wanted to use AI to “deduce” — or draw conclusions from this data to help employees perform day-to-day tasks — the cost will be even higher.
Companies that already understand this dynamic, according to Michael Dell, are considering “jumping ship” from public cloud storage purely and simply due to costs.
Jeff Boudreau, President of AI Strategy at Dell on the left and Matt Baker, Vice President of AI Strategy at Dell on the right (Imagel: Léo Müller)
The issue of data security is another important factor weighing against processing data and via AI in public cloud services. If you have proprietary source code in your company, confidential customer data or even something with intellectual property (films, series, music, schematics for a new product), keeping this in the cloud would not be a good practice.
“Some of our most sensitive intellectual property information needs to be kept only locally and never sent to the cloud,” warned Jeff Boudreau, president of AI strategy at Dell.
Although this “local versus cloud” dilemma is an issue for a company’s managers to resolve, those who work on the cutting edge need to keep in mind that throwing anything at ChatGPT may not be a good idea, depending on the type of work you do.
“These public Gen AI solutions are great for the end consumer, but you need to remember that these models were not trained on your company’s practices and data,” Matt Baker, vice president of AI strategy at Dell, warned managers of IT in the corporate world.
“If you are working with publicly available general knowledge, use a public model. But if you want to create effective technology for your organization, you need to leverage your own data, your knowledge and information. And if you decide to do all of this, why not run your own LLM too? It will be much cheaper”, recalled Baker.
The executive also used the sustainability card, which should appeal to companies with goals to reduce their carbon footprint generated by their operations. After all, keeping data in the cloud or on your own servers and “making them work” for the company involves considerable electricity use. And in many locations around the world, this resource is generated by burning fossil fuels to this day.
“Why run a 30 billion parameter LLM to get answers about something I have all the information about locally? Why not run a 7 billion model that is more efficient and uses less energy and gets the same response?”, provoked Baker.
With all of this, it’s clear that you really shouldn’t be using ChatGPT, Gemini or Copilot to carry out your daily tasks, but management is the one who needs to decide which tool to use and whether it’s really worth using one, not the employee.
Gen AI and its potential to create a new industrial revolution
“The first factories were mills. Water or wind turned a wheel, and that wheel did some work. Everything you could automate had to go through the wheel. With the advent of electricity, what did they do? They used electricity to turn the It took us many years to decide to bring electricity everywhere and automate many more tasks with intelligent and specialized machines that we could plug into the socket. Don’t use AI to spin the wheel, but rather reimagine and reinvent your organizations to master this hyper-intelligence.”
This somewhat prophetic speech by Michael Dell when describing the state of artificial intelligence with an analogy may not appeal to everyone, but it brings an important warning to large companies in particular. As they are the largest holders of data, they are also in the best position to start training their own AIs.
Providing this to an overworked workforce after waves of layoffs in various sectors of the economy around the world can generate gains in productivity at a relatively low cost. In short, the idea is that it will be possible to do more with the same workforce that companies already have and, consequently, stand out compared to their competitors.
And bringing the conversation to the worker context, those who have more creative positions or skills that are more difficult to find on the market have a very low chance of being impacted by the use of AI on a large scale. However, anyone who does mechanical work will certainly have to reinvent themselves sooner or later.
Will an AI steal your job?
This issue of the impact of AI on the job market is so present that it becomes evident wherever it is adopted. FreeGameGuide spoke with Yan Chen, studio architect at Kennedy Miller Mitchell (KMM) — producer of the film Furiosa: A Mad Max Saga — to understand how George Miller’s film used AI in its production, and it became clear that the objective of adoption technology is about doing more with less.
This is an excellent approach for companies or projects with limited resources. With an AI trained and operated with Dell infrastructure, KMM was able to save a lot in the editing and visual effects creation process.
“Traditionally, you finish visual effects work when you run out of time or run out of money. This is because the specialized labor for this is very expensive and it takes time to make everything perfect. Therefore, the director is unlikely to be satisfied with the final product. But this is the first one where George Miller is 99.99% satisfied with the look of the film,” Chen said.
Chen also highlighted that, with the use of Gen AI in Hollywood, studios will end up being able to produce many more films in the same space of time. In other words, this would be the materialization of the transformative productivity gains that Michael Dell referred to in his opening keynote at Dell Technologies World.
In the case of the North American film industry, however, this will imply a distribution problem for all this new content that the studios will have to resolve.
“Gen AI is really an industrial revolution because it will allow us to go from about 30 films a year in Hollywood to something like 300 or maybe 400. But this requires a change in our distribution model: will we even want to go to the cinema to see all these films?”, concluded Chen.
The case is that LLMs, and Gen IA as a whole, have undeniable transformative potential at this point. However, we still don’t know how to deal with this entire transformation, and whoever manages to solve this enigma first will probably have a lot of money in their pockets in the coming decades.
This applies to both the businessman and the worker because, even if Gen AI evangelists argue that technology will not steal anyone’s work, it is innocent to think that work itself will not be transformed. Or do you know someone who works as a telephone operator in 2024?
