Trends in e thic al marke t ing




Download 5,31 Mb.
Pdf ko'rish
bet12/14
Sana04.06.2024
Hajmi5,31 Mb.
#260281
1   ...   6   7   8   9   10   11   12   13   14
Bog'liq
sf-418-trends-in-ethical-marketing-guide-081623

The Making of 
Responsible Marketing 
at Salesforce
72%
High Performing Marketers
61%
Underperforming Marketers
Ability to analyze marketing performance in real time
2
:
72%
Consumers will remain loyal to companies that
1
:
Deliver faster service
65%
Offer a more personalized experience
of consumers say 
they will reassess their 
budget over the next 
12 months as they seek 
more personalized 
experiences
1
81%
1 State of the Connected Customer, Salesforce 
2 8th Edition State of Marketing, Salesforce


12
TRENDS IN E THIC AL MARKE T ING
SALE SFORCE
The Future of 
Marketing & AI?
When it comes to specific technologies like 
AI, we support accountability through publicly 
available commitments like the model cards 
mentioned earlier and our Generative AI 
guidelines for responsible development.


1 3
TRENDS IN E THIC AL MARKE T ING
SALE SFORCE
Generative AI & Responsibility
Pairing the power of technology with Salesforce’s passion for trust, our principles for responsible 
generative AI development aim to keep our values at the core of our approach to innovation.
Accuracy: 
We need to deliver verifiable results 
that balance accuracy, precision, and recall in the 
models by enabling customers to train models on 
their own data. We should communicate when there 
is uncertainty about the veracity of the AI’s response 
and enable users to validate these responses. This 
can be done by citing sources, explainability of why 
the AI gave the responses it did (e.g., chain-of-thought 
prompts), highlighting areas to double-check (e.g., 
statistics, recommendations, dates), and creating 
guardrails that prevent some tasks from being fully 
automated (e.g., launch code into a production 
environment without a human review).
Safety:
As with all of our AI models, we should 
make every effort to mitigate bias, toxicity, and 
harmful output by conducting bias, explainability, 
and robustness assessments, and red teaming. 
We must also protect the privacy of any personally 
identifying information (PII) present in data used for 
training and create guardrails to prevent additional 
harm (e.g., force publishing code to a sandbox rather 
than automatically pushing to production).

Download 5,31 Mb.
1   ...   6   7   8   9   10   11   12   13   14




Download 5,31 Mb.
Pdf ko'rish