AI and the Tech Industry

 Before I say anything on this, I want to be clear: I didn't get ChatGPT to write this blog article...  


AI in general (and ChatGPT specifically) have revolutionised content creation by seeming to understand complex topics and express them in perfect English (or Spanish, or Lithuanian, or whatever language you want...).  

What does this do for us?

The topics ChatGPT understands are things that have already been written about on the internet.  The beauty of ChatGPT is that it acts like a super intelligent search engine, giving you the answer to a question instead of sending you loads of hyperlinks to research.

A great use in the tech industry is to ask it to compare two products using various criteria.

However, it soon becomes unstuck if you ask very specific questions, like "will X work with Y".  It can produce misleading content or stay very generalised when you want a specific answer. 

So how should we treat it?

My view is that any content ChatGPT creates should be viewed as a summary for your own personal benefit.  If you treat is as something to post about on a blog, you are simply adding more spammy web pages to the topic, which it'll consume again in time, and this feedback loop will ultimately turn any content on the topic into grey mush.

Does this mean that it will never contribute meaningfully?

No, future AI tools will of course have the capacity to contribute more meaningfully.  The capability exists but needs to be more "joined up".  For example an AI tool could build software, analyse user behaviour, write manuals on use of the software, and so on.  But this brings up a bigger question...

Should we give AI tools 'agency'?

Can we be safe giving AI a range of abilities that are able to grow as they explore the power of the internet, and the ability to make decisions on how to use them?  In fact a tool called "ChatGPT God Mode" already exists to do this.




Comments

Post a Comment

Popular posts from this blog

Microsoft CIAM debuts at Build 2023

Image
For a long time we've been aware of Microsoft's efforts towards converging Azure Active Directory B2C and Azure Active Directory. They are making this change in order to provide a single stable product for the back end portal and other aspects of these identity services.  New developments will see a 'CIAM' product sitting on top of AAD - this replaces (or will eventually replace) B2C.   CIAM will also provide better experiences for developers hoping to build user journeys (sign in, sign up, etc.). Microsoft continues to support AAD B2C.  At time of writing it provides a fuller feature set.  However, B2C as-is will eventually retire, and migration paths will be important. The CIAM solution will offer more back-end management features, and an event-driven programming paradigm for user journeys. I will make more information available as and when!
Read more

LinkedIn now uses Verified ID for Employee verification

Image
  Microsoft Verified ID is a digital identity verification solution that enables individuals to obtain digital "credentials" from organisations that know them, or more accurately, that know some aspect of their identity.  For example, their name, their educational qualifications, or in this case, their employment status.  These credentials can then be shared with third-party organizations, such as employers or educational institutions, in a secure and trusted manner.  LinkedIn, which is owned by Microsoft, has integrated with Microsoft Verified ID to allow LinkedIn members to verify their employment status. Using this new integration, employers can create a custom process for verifying employees, and employees, once they have done this, can present the credential to LinkedIn in order to obtain the new green "check mark" on their LinkedIn profile. Currently this requires the LinkedIn app - at time of writing there is no announcement yet about migrating this feature t
Read more

Keeping safe beyond sign-in

Image
As The Register reports today , Microsoft briefly exposed a plethora of passwords and other secrets due to a bulk upload of data to GitHub recently. The data was a set of AI training data - data used to teach AI models how to understand the world around them. It makes no sense having the best locks on your front door if you leave the back door open... so what can we learn from this episode? Protecting passwords and other sensitive information from exposure to the wider world is essential for maintaining your online security. Here are some measures you can take: Use Strong, Unique Passwords: Create strong, complex passwords that are difficult to guess. Use a combination of uppercase and lowercase letters, numbers, and special characters. Avoid common dictionary words or phrases. Ensure your passwords are at least 12-16 characters long. Use a passphrase, which is a sequence of random words or a sentence that's easy to remember but hard to crack. Use a Password Manager: A password man
Read more