Business, operational and ethical concerns
Large Language Models for Business
Iason Prassides
Content Developer, DataCamp
Transparency
Transparency around LLMs
Businesses must know how an LLM gets to a response
Essential to create trust
Accountability
If LLM's recommendations don't work
Who is responsible?
Software developers that designed the model?
The company that deployed it?
The risks with LLMs
Inherent risk of propagating misinformation
Repercussions for businesses including:
Basing decisions on incorrect data
Impact bottom line
Damage reputation
Danger of LLMs being weaponized:
Deceive, manipulate, mislead stakeholders
Cause reputational and operational risks
LLMs and the environment
Computational power used to train LLMs requires significant energy
Creates large carbon footprint
Businesses must understand and mitigate environmental impacts
How to build LLMs?
Businesses have pragmatic considerations to weigh
Build in-house LLMs or use third-party providers?
Technological resources
In-house
Significant resources required - computational infrastructure and data
Can investment be made? Money and time?
Third-party
Provide immediate access and scalability
No development overheads
Personalized LLMs
In-house
Tailor LLMs to specific needs
Fine-tuned models based on company data
Third-party
Sophisticated performance
More generic models, unless customized
Updates and maintenance
In-house
Maintenance ensures LLMs remain effective and secure
Ongoing cost for business
Third-party
Companies not responsible
Provider handles updates and maintenance
Handling data
In-house
Maintain control over data
Important when using sensitive data
Third-party
Could require sharing data
Possible security and privacy risks
Cost efficiency
In-house
Balance recurring costs to ensure cost effectiveness
Start up costs are high
Third-party
Recurring costs
Startup costs are low
The choice is yours!
Let's practice!
Large Language Models for Business
Preparing Video For Download...