Discover more from The Overfit
Is OpenAI... really Open?
With the "secretive" launch of GPT-4, many questions are raised.
OpenAI is a company dedicated to researching AI, and developing it responsibly. Initially, the “Open” from its name stems from “open-source,” meaning that their research would be published and open for the public, researchers, and so on build upon to evolve the field as a whole and provide transparency.
Note: GPT-4 is a new release of OpenAI that can handle multimodal data — text, image, audio, and video.
However, they might not be so transparent:
The technical report they released includes the following phrase:
Given both the competitive landscape and the safety implications of large-scale models like GPT-4, this report contains no further details about the architecture (including model size), hardware, training compute, dataset construcction, training method, or similar.
Therefore, they decided to make more things “confidential” in the name of “competitiveness” and “safety.”
I think it’s okay for a company to approach the field with a more “closed” mindset. For example, Apple is very secretive about its AI Research (or at least way more than Google, Meta, etc.). However, it’s strange for a company that includes “Open” in its name to release so few details about a new model.
Was it “Open” until now?
There were other instances in which the “openness” of the company was questioned. For example, when they announced an exclusive partnership with Microsoft.
OpenAI, founded in 2015, was exclusively a “Non-Profit” organization, backed with $1 billion to fund their research efforts.
In 2019, OpenAI transitioned to a “capped” for-profit model. Essentially, they would keep the profits capped at 100x the investments [big cap, huh?]. The argument presented here was based on finding venture capital to fund their research would be facilitated, given that a great return would be possible.
Even with this change, they have divided the company into OpenAI Inc. (non-profit) and OpenAI LP (for-profit). Still, given the changes, even if the company still wants to develop AI responsibly and “openly,” it’s clear that they might have to compromise that to achieve profitability, precisely what is happening now.
I expect this move to be openly (haha) criticized, and it will most likely be a challenge for OpenAI to decide how “open” they want to be.
Let’s see if they will keep on this “closed” path or work more openly on other fronts! 🤔