By Steve Cracknell, Founder Insig Ai
9 March 2023
Generative AI: The game-changing access to advanced technology
The hype around Generative AI (GenAI) is real, and the advancements we will start to see with this technology being made widely accessible are going to be exponential.
AI has always been veiled within its infamous ‘black box’, with most people being concerned and dubious about what is going on under the hood.
With GenAI and specifically ChatGPT however, you see its behaviour unfold in front of your eyes and you can question its logic and challenge its responses. In addition to being able to ask complex questions, you can also ask it to perform useful sorting, summarising and explanation tasks of your own data.
Contextual data and text-based algorithms like ChatGPT are insanely difficult to build and train (let alone use) and have historically only been available to a very small and specialist audience of proverbial ‘rocket scientists’.
The excitement being seen in the market around GenAI is a testament to making complex technology easy to use and widely accessible. In fact, I would go as far as suggesting the rebranding of AI from ‘Artificial Intelligence’ to ‘Accessible Intelligence’.
This really is going to be a game changer for businesses.
How we make GenAI complement our own technology
As a technology company that develops our own machine learning algorithms, NLP (natural language processing) classifiers and AI tools to help asset managers look for better and faster ways to find alpha and attract capital with high quality ESG credentials, the most interesting thing about GenAI is the application of these sophisticated contextual models to our own datasets.
We have over 100 million corporate ESG disclosure sentences in our database, covering 4k major global companies and 160k unique documents. With all our data stored in a machine-readable format, the utility of ChatGPT and other GenAI tools is vast.
We have solved the data collection, transformation out of PDF and storage problem for corporate ESG disclosures. Our 15 NLP classifiers help group the disclosures into more manageable and consumable topics, e.g. climate change, water, supply chain management and workforce.
Applying the contextual data management and sorting capabilities of ChatGPT to our filtered dataset helps to deliver the ‘last mile’ when it comes to providing and packaging actionable insights out of our data for our clients.
Scrutinising huge volumes of corporate disclosure
Looking at this more practically, here are some examples.
At the most basic level, you can ask ChatGPT to summarise the key points from a body of text. The Chairmans’ Statement in an Annual Report is a good example. Once you get the summary back, you can ask ChatGPT to split the sentences into positive and negative sentiment blocks.
ChatGPT data sorting and summarising capability
Another functional use case is asking ChatGPT to return references to “all net zero targets being set in the future by a company, but do so in a tabular format, splitting text from dates”.
Insig.AI NLP classifiers returns sentences about ‘Climate Change’
ChatGPT data formatting for easy reference
From a productivity perspective, GenAI will help remove many of the mundane tasks required to surface and make sense of the data you care about most.
These examples still only scratch the surface of what is possible. With regulation starting to bite on requirements around Article 8 and 9 funds, tracking the SFDR Principal Adverse Impacts (PIA) indicators is front of mind for many of our clients.
“Summarise the key SFDR disclosures that talk about activities negatively affecting biodiversity sensitive sites. Return the results in a table that can be exported as a .CSV.”
Insig.AI search quickly narrows core analysis data set
ChatGPT smart summarizing and data formatting
How to make ChatGPT work for you
There are a few fundamental points to consider to when using ChatGPT, especially if you want to use it at scale:
The quality of the question matters: You pay for every token (word) that the machine reads. If you ask a dumb question, you will get a dumb answer…and pay for it!
Being suitably skilled to interpret an AI generated answer: If you have zero domain expertise on your topic of choice, you won’t be able to differentiate a good response from a bad one.
ChatGPT does not currently access a real-time source of up-to-date information: Having an accessible repository of machine-readable data will allow you to interrogate and surface current content most relevant to you.
Knowing how to integrate this new technology into your current workflow: To use ChatGPT at scale and affordably requires experienced DevOps engineers and extensive cloud expertise.
These points will ultimately separate ‘hobby’ users of ChatGPT from those who can truly take business advantage of it.
For Insig.AI, we see the advancement of this type of text-based analysis being a perfect overlay to our vast repository of machine-readable ESG and corporate disclosure data.
Combining the deep contextual understanding that ChatGPT brings with our highly focused ESG NLP classifiers will be differential in surfacing the most relevant ESG disclosures to our clients and increase analysis potential and productivity.
Комментарии