Coincidentally or not, Grammarly raised $200 million during Academic Writing Month. This is a lot, a very much lot! They raised $400 million to date, and are valued at $13 billionš¤Æ. The company runs a freemium model, but thatās not how it started. They were selling to universities first and already had hefty revenues before they moved to charge individual users. That helped them expand and reach profitability.
A similar product focused on academic writing and the jargon of scholarly communications is Writefull, founded in 2014 and part of the Digital Science portfolio. Over the summer, Writefull released a new version that incorporates some cool algorithms. It can both correct errors and recommend the phraseology that youād use when writing a paper, for example relating to the methods or the results. This is an absolute saviour for many non-English speaking researchers. Broken English can reduce the perceived validity of arguments.
Whilst writing complete works for the general public is difficult for an AI because it needs to achieve coherence, an academic article is bound by more structure and jargon. Which means that an AI could write an academic article more easily (in fact there is a whole industry of fake articles).
Employing GPTs is somewhat more practical for writing abstracts. For example, this automated abstract writer based on GPT-2 and trained on PubMed articles is not that bad.
Two other AI writers that I tried (articoolo and AI-Writer) were not as impressive, and in multiple instances, it was clear that more than a phrase came from somewhere else, plagiarism some?š¤
šAutomating parts of the writing process has been around since before the GPTs. Already in 2011, Bloomberg used algorithms that auto-generated short pieces on company earning releases; it had to be super fast, and that was the only way. I wasnāt surprised to read that at least a third of Bloomberg news articles were written by some form of AI in 2019. There are similar reports for other news publishers.
š¦¾Writing code is also very structured, it has rules, itās perfect for automation. About 30% of code in GitHub is automatically written by Copilot.
āļøRecently, in AI for writing:
Anyword is optimizing marketing messages; raised $21M.
Copy.ai uses GPT-3 to help with business copywriting; raised $13.9M in 2021.
Writer is another writing tool for businesses; raised $21M.
Katteb generates content for bloggers; launched in September 2021.
Craftly.ai generates content for products and services; based on GPT-3 and also launched earlier this year.
Symbl.ai is next level, itās a developer tool to integrate āconversation intelligenceā in their apps; raised $17M.
A bit adjacent to this category - Verbit, the US-based tool for transcription, raised $250M! They raised about the same amount previously.
If you want to dig into more tools automating the scholarly publishing process, have a look at this incredible radar from PubTech š¤©.