Tuesday, September 19, 2023

Deciding on Generative AI: A Risk Analysis Approach

 Deciding whether to let ChatGPT, or other Generative AI tools, loose at your company is difficult đŸ€”




👍 On the one hand, you read that 80% of Fortune 500 companies have adopted ChatGPT. FOMO.

đŸš« On the other hand, you read that 75% of organizations worldwide are set to ban ChatGPT. Better not use it


đŸ€· This boils down to the uncomfortable answer that it's up to you to make the call.

📋 A great way of making tough decisions is to consult the good old risk analysis template to see what the major risks are for your business and address them.

📊 McKinsey & Co. conducted a survey on the most cited risks of generative AI adoption. This can serve as a great starting point to kick off the discussions and decide what categories to include in your risk analysis.

What strikes me about the categories is how performing the risk analysis is a full-company exercise (or all of your brain cells if you are a solo entrepreneur). It stretches from pure IT aspects such as cybersecurity risks, to organizational reputation, to the risk of your staff being worried about displacement.

📄 To help you along the way, I've put the risks into a risk analysis template with a few examples for each of the categories. Reach out to me if you want the Excel file.

⚠ NOTE: The risk categories and examples are just for sparking your thinking; you are ultimately responsible for your own risk analysis.

Do you think a risk analysis could clarify you thinking about using generative AI?



Thursday, September 14, 2023

Nu finns allt för att vÄga prova Generativ AI, eller?

 Ă„r “Copilot Copyright Commitment” frĂ„n Microsoft det som krĂ€vs för att alla slutligen ska vĂ„ga experimentera brett med generativ AI och rulla ut skarpt?



Generativ AI förvĂ€ntas öka vĂ„r produktivitet rejĂ€lt, McKinsey uppskattar ökningens vĂ€rde till 4.2-6.4 trillioner dollar per Ă„r. JĂ€ttemycket pengar. Generativ AI kommer bli viktigt i dom flesta organisationer, kanske pĂ„ samma sĂ€tt som internet eller rent av elektricitet âšĄïž

Men, vÄgar man testa?

Det Àr mÄnga utmaningar som behöver hanteras:

♄ Vi mĂ„ste vĂ€rna om vĂ„ra anstĂ€llda och inte skapa panik för att bli av med jobbet.

đŸ•”ïž Vi behöver lita pĂ„ att AI systemen inte exponerar vĂ„ra hemligheter, eller utsĂ€tter oss för IT-sĂ€kerhetsrisker.

đŸ‘©â€âš–ïž Vi mĂ„ste vĂ„ga anvĂ€nda det som kommer ur AI verktygen utan att riskera att bli stĂ€mda för IP intrĂ„ng.

Tar man ett stort bolag som Microsoft har dom förstÄs adresserat allt detta och bitvis kommunicerat ut det:

♄ Tidigt började man anvĂ€nda “Co-”, Copilot, Co-create, o.s.v., för att var tydliga med att vi mĂ€nniskor sitter i förarsĂ€tet och AI-verktygen hjĂ€lper till oss bĂ€ttre. Inte ersĂ€tta.

🔒 Man Ă€r tydlig med att all insamlad data Ă€r anonymiserad och aggregerad, vilket sĂ€kerstĂ€ller att enskilda anvĂ€ndare eller organisationer inte kan identifieras. Samma rigorösa sĂ€kerhetsstandarder gĂ€ller för Copilot som för Mirosoft 365.

📑 För nĂ„gon vecka sedan slĂ€pptes Copilot Copyright Commitment som ska göra oss trygga i att IP som genereras tillhör skaparen. Man strĂ€cker sig sĂ„ lĂ„ngt att till och med tĂ€cka upp för eventuella stĂ€mningar och skriver i sin blogg:

“Copilot Copyright Commitment. As customers ask whether they can use Microsoft’s Copilot services and the output they generate without worrying about copyright claims, we are providing a straightforward answer: yes, you can, and if you are challenged on copyright grounds, we will assume responsibility for the potential legal risks involved.”

Är Copilot Copyright Commitment det sista som behövdes för att ni ska börja experimentera med generativ AI pĂ„ riktigt?


Referenser

https://blogs.microsoft.com/wp-content/uploads/prod/sites/5/2023/07/Microsoft-Voluntary-Commitments-July-21-2023.pdf

Data, Privacy, and Security for Microsoft 365 Copilot - Deploy Office | Microsoft Learn

Microsoft announces new Copilot Copyright Commitment for customers - Microsoft On the Issues

Economic potential of generative AI | McKinsey