Image by editor
# introduction
what to do thing Do they have things in common like electricity, wheat, mobile phones, and the internet? Probably, for many, they have become the so-called “.” merchandisethat is, “a resource or commodity that is completely or substantially substitutable (or considered simply essential to modern lifestyles)”.
Most of the examples mentioned became commodities at certain turning points in history. For example, in the late ’90s, at least as far as I can remember, cell phones were still an up-and-coming phenomenon. But in this century, who can imagine life without it?
In this article, we fast forward to the present day and analyze a series of objective facts about the wave of language modeling technology that has swept into our lives over the past three to four years. I would like to add some personal thoughts regarding the following questions. Are language models a new commodity that we can’t live without?
# check the facts
Just a few years ago, advanced language models were considered a luxury technology asset, but now they are a ubiquitous solution that many organizations cannot afford to skip.
There are several facts about current market realities that explain this widespread accessibility to language models.
- Decreasing costs: This may sound counterintuitive in today’s global climate where prices for just about everything are rising, but one exception to this norm is the cost of “raw intelligence” solutions. An example is the cost of processing 1 million tokens (approximately 750,000 words) on the Frontier model. What used to cost tens of dollars even a few years ago can now cost tens of dollars. cent.
- Free Access Revolution: The open weight model helped break down the exclusivity barrier. Meta’s family of language models, such as Llama and Mistral, have demonstrated that they can match or outperform many commercial language alternatives based on public benchmarks.
- Zero cost: Any user can now download a free language modeling tool like Ollama and run powerful models locally on their machine. This completely eliminates the need for paid subscriptions or API quota usage, as well as any dependence on third-party services. As a result, access to AI as a basic, free resource has become the new norm.
“For example, in the late ’90s, cell phones were something of a budding new thing — at least as far as I can remember. But in this century, who can imagine life without them?”
# Explore fact-based opinions
Basic AI capabilities may have become a commodity, but having a model with a “personality of its own” that performs complex, detailed tasks to the best of its ability still doesn’t seem like it. Although many underlying models (which are foundational from an upstream general-purpose perspective and represent an initial, extensive pre-training stage before the model is adapted to a specific task) can generate text responses or codes for free, there are still noticeable differences between natural human narratives and the conversational style of the models, and the words and expressions they use can sometimes be somewhat robotic and predictable. Many of the outputs produced by these models still require final tuning, which is what makes them different in different areas.
On a different note, many users don’t invest in the model itself, but rather in the experience that comes with it. It sounds great to have a free text generation engine on our computers, but companies need to charge for improved solutions that allow their models to interact with documents, code, or workflows in specific ways. We are still relatively far from the point where everyone is fully accepting of delegating this task to a (often paid) language modeling solution rather than doing it themselves.
# give a verdict
Based on the facts that reveal how the role of language models and access to them has evolved in recent years and become almost free, it can be argued that these factors have propelled such models into the new commodity of this decade. However, there are still aspects that are not available to everyone, such as reliability, fully guaranteed privacy, and adaptation to niche application areas (such as medical or legal reasoning). premium Products, and the term “product” in the context of language models has become somewhat controversial.
Ivan Palomares Carrascosa I am a leader, writer, speaker, and advisor in AI, machine learning, deep learning, and LLM. He trains and coaches others to leverage AI in the real world.
