| |
To date OutScribe has had a policy of not editing AI generated transcripts because they were so inaccurate it was actually quicker to transcribe from scratch. With the improvement of AI technology we have now reviewed that policy and will edit reasonable quality transcripts. The usual $/min cost will be discounted once the quality of the transcript has been checked. Microsoft Teams transcripts have been fairly good, with Zoom catching up quickly. Does OutScribe see this improvement as a threat or opportunity? We certainly don’t see it as a threat but rather, a useful tool that can be utilised for some recordings. Not all recordings or work types are suited but we are intending to incorporate an AI option into our suite of services in the future for clients who feel they may be in a position to benefit. Before we select and integrate any AI system into ours, not only must it deliver accuracy and efficacy but even of greater importance to us it must provide proof and alleviate concerns around AI transparency, privacy and security. In the current AI arms race between tech giants there remains significant risks and ethical dilemmas. For example, just a few months ago Zoom was caught giving themselves permission to use client meetings for AI training and it's unclear if Microsoft, who is the largest stake holder of OpenAI (ChatGPT), trains AI from its considerable resources, ie. Sharepoint, Office, Windows OS, Skype, Teams, Xbox or any of its other 130 products. The power of AI now more than ever increases the ability to attribute previously anonymous data to the actual individuals, with the obvious temptations for corporate and government entities to then leverage this in marketing, societal control, insurance exposure reduction, etc. In short, before bringing AI into the OutScribe fold we want to satisfy ourselves that we can deliver the benefits of AI without compromising or eroding the privacy and security standards that OutScribe has become known for.
|
|
