17 May The Clone Wars: ChatGPT vs. the Clones
ChatGPT’s beguiling ability to provide convincing answers to an incredible range of queries has caused a cultural phenomena that’s reverberating around the world. It’s also kicked-off a deadly serious war among tech companies to capture #mindshare first, and #marketshare second, which I explained in a previous post The war of the Machines: In the great battle of LLMs, what side should enterprises be on? This war features startups, tech companies, even nations, from “prince to pauper,” as the saying goes, announcing LLM models. Now following these announcements — barely a few months — the race has shifted to the next phase of industrialization of AI, the rise of the clones.
Industrialization – the next phase of AI
Whenever a new technology arrives, beneath the hype, a multiyear effort to “industrialize” occurs; the end product evolves, modifies, and gets refined in a constant improvement cycle. The best recent example is the smartphone: it took about 20 years for the brick Gordon Gekko used in Wall Street to become the sleek smartphone we use today with incredible finishes and all the new advances in glass manufacturing and technology.
The Evolution of the Mobile Phone
What we see happening in the industrialization of AI, is the same phenomenon occuring, but in months, not years. The industrialization curve is now so compressed, “the clones” are being launched almost every week.
Open Source: The Rebel Alliance
Hugging Face: The open-source startup recently launched a ChatGPT clone, based on LAION’s LLaMa-based ‘Open Assistant’. The service provides an easy way for people to access a ChatGPT-like system, albeit based on openly shared underlying models. Hugging face continues to snag benefactors including IBM’s newly announced partnership as well as a partnership with amazon.
The rest of clones are in formation: Stability AI open-sourced it’s advanced image-generation technology successfully, and is now working on an open source competitor to ChatGPT. “We are a few months from release,” says Emad Mostaque, Stability’s CEO. Anthropic, Cohere, and AI21 are also working on proprietary clones to OpenAI’s. Databricks made a big splash with its Dolly 2.0 release, reportedly the first open source, instruction-following LLM for commercial use that has been fine-tuned on a human-generated data set.
OpenAI’s Sam Altman getting ready for battle in the great clone war – image by Midjourney
What’s an Enterprise supposed to do?
Enterprise buyers are flummoxed with this frenzy of announcements and watching a rare phenomenon in software which is common behind the covers in silicon valley. For the first time, iterating, rebuilding, improving, and the refining of LLM models is happening in plain sight spurred by murmurs of AGI and interspersed with cries for regulation.
Enterprises have two choices:
Dive in? Some enterprises will need to dive in and try use cases with a trusted supplier and keep Iterating with guardrails around data privacy and security. The advancements in productivity automation and progress is simply too important to be left behind in industries like customer service, IT and HR automation.
Wait and watch? For those for whom data privacy and liability from errors is too big a business risk (like healthcare and financial services) there will be a wait and watch attitude till the dust settles. First regulatory capture needs to be codified and more robust consumption patterns should be identified, which will normalize commercial considerations.
Irrespective of the industry, the Clone Wars is heralding a new era in technology where speed and innovation is being witnessed in plain sight. Multiple Jedi Knights will be leading the war and spectacular successes and failures will be norm. However the advances in business efficiency and human computer interaction will simply be too significant to ignore.