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

The word "Truth" and "Trust" being written out by wooden blocks

Large Language Models for Business

Accountability

Image showing wooden figures in a circle, with one different colored figure in the middle

  • If LLM's recommendations don't work
    • Who is responsible?
  • Software developers that designed the model?
  • The company that deployed it?
Large Language Models for Business

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

Cartoon showing a man standing on blocks spelling out "Risk" while measuring the height of the risk

Large Language Models for Business

LLMs and the environment

  • Computational power used to train LLMs requires significant energy
  • Creates large carbon footprint
  • Businesses must understand and mitigate environmental impacts

A robotic hand holding the planet Earth, with a tree growing from the top

Large Language Models for Business

How to build LLMs?

  • Businesses have pragmatic considerations to weigh
  • Build in-house LLMs or use third-party providers?
Large Language Models for Business

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

Cartoon image of a man stacking coins while standing on a giant alarm clock

Image of a notebook showing decreasing operating expenses, with a coffee mug and a man's arms also showing

Large Language Models for Business

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

Image showing the word "customization" being touched by a human finger

Image showing generic products stacked in a supermarket, with the products different shades of the same color

Large Language Models for Business

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

Cartoon image of a woman on her phone, with images of maintenance and technology updates in the background

Image of the words "Third Party Service" on a chart clipboard

Large Language Models for Business

Handling data

In-house

  • Maintain control over data
  • Important when using sensitive data

Third-party

  • Could require sharing data
  • Possible security and privacy risks

Cartoon image showing a man using a laptop while shaking another man's hand while he is in the laptop screen

Cartoon images showing different characters stealing and obtaining information from a laptop

Large Language Models for Business

Cost efficiency

In-house

  • Balance recurring costs to ensure cost effectiveness
  • Start up costs are high

Third-party

  • Recurring costs
  • Startup costs are low

Cartoon image showing a man holding a plug and socket that will connect money with a lightbulb

Cartoon image of a man carrying a card in front of a specific date on a calendar

Large Language Models for Business

The choice is yours!

A signpost showing two possible directions - outsource or inhouse

Large Language Models for Business

Let's practice!

Large Language Models for Business

Preparing Video For Download...