Garbage In Gospel Out (GIGO) was a humorous, ironic statement from the 1960s observing the reverence directed towards mainframe computer printouts. The acronym was applied as confidence was given to computer-printed reports over statements of experience or non-quantified claims in business discussions.
GIGO today might also be an apt acronym for GPT In Gospel Out from the ‘ask me anything’ sessions with OpenAI’s ChatGPT (Generative Pretrained Transformer Versions 3 and 4). EkaLore feels that a little skepticism is in order.
It’s not just ChatGPT. Several LLM (Language Model for a Large-scale content) systems, including Google Bard, are being released as pre-trained models with billions, or even trillions, of parameters that use human structured inputs and natural language processing to generate responses. However, there are limitations to the current approach when defining "ethical," as the rules and progress are still being refined.
History shows that the early use of tools by highly productive individuals can produce extraordinary results in short periods of time. The sustaining effects of integrated tool usage are likely to be seen in end-user applications of the LLM/chat-style tools as developers and users become more conversant with their advantages and limits.
Experience exists with ‘search’ applications and deep knowledge bases like Lexis/Nexis, domain knowledge-intensive tool capabilities like Wolfram Alpha, and language processing in the evolving healthcare solutions from Merative (formerly IBM Watson Health). These mature tools have been in use for over a decade and continue to evolve. Applications of these tools in widely used applications include search (Bing, Siri, Alexa), spreadsheets (Excel), medicine (Watson), and writing (Lexis for Microsoft Office). Many of these applications are/have evolved past first-generation into tools where the embedded or incorporated features are just part of the user functionality.
ChatGPT and other LLMs are advanced tools and will provide substantial value. As with all tools, users will need time to develop their expertise and derive value from them.
EkaLore writes about AI/Big Data/Data Science topics on our blog at www.ekalore.com/bad-project-blog
If you’d like to book an initial free consultation with a senior analyst send us a note at sales@ekalore.com
Comentários