The American software publisher Databricks has launched Dolly 2.0. It would be the first open source language model for commercial use, likely to form the basis of free or open source competitors to chatbots such as ChatGPT.
Dolly 2.0 is a text generator, a ‘large language model’ (llm) capable of driving chatbots and summarizing texts or executing search instructions. In terms of functions, Dolly 2.0 therefore strongly resembles other llms released or announced in recent months, of which ChatGPT is probably the best known. What’s special regarding this though is that Dolly 2.0 is open source and independent developers and companies can license the service for free, in order to incorporate it into commercial products.
The AI is a sequel to the first Dolly, which was released in March this year. The company behind the llm is called Databricks, a professional software publisher specializing in Apache Sparks. “We want to bring more open and transparent language models to market, because we want companies to be able to use AI to create, train and own their own chatbots and apps,” Ali Ghodsi, CEO of Databricks, tells the technology site TechCrunch.
No restrictions
What is amazing is that Dolly 2.0 is therefore the first generative AI model to be launched without any restrictions in the market. Companies that want to work with ChatGPT, for example, pay for the API and also often have to share data. With an open source model, the system should prove usable by more companies.
That said, Dolly 2.0 seems to have not only similar advantages, but also characteristic disadvantages of other llms. This is how the algorithm would mostly speak English and would tend to imagine facts. He would sometimes also be offensive in his expressions. All these problems had already been noted in other linguistic models.
The development of generative AI has literally taken off in recent months, as several technological giants are trying to position themselves in this market. Linguistic models, however, are also subject to criticism. Thus, questions arise regarding the sources from which these models are formed, and/or regarding whether the authors of these sources have given their consent for their use. International experts have also started a petition asking that the development of AI be suspended for a while. In particular, they worry regarding the economic effects of technology, as well as the dangers of large-scale misinformation, should these chatbots flood our internet.