I will unveil the findings of an exhaustive study on the current landscape of AI in business, its market dynamics, and the associated risks. This research, rooted in the Natural Language Processing (NLP) analysis of media discourse, utilised leading language models (LLMs) and drew from a variety of data sources. Beyond the empirical picture, I will dive into an analysis of the various risk levels inherent to the AI business. Additionally, I will provide insights into the potential direction of the AI market in the near to medium-term future. The primary objective is to clarify our present understanding and control of the AI business, while also highlighting the uncharted areas we might encounter in the future.

1. DECIPHERING THE
AI LANDSCAPE:
BUSINESS, MARKET
TRENDS, AND
EMERGING RISKS Goran S. Milovanović
Lead Data Scientist/Owner

2. On the nature of revolution

3. Natural Intelligence

4. System1|System2

5. HOSPITAL
NURSE
DOCTOR
DOCTOR
PHILOSOPHY
UNIVERSITY

6. System1 == Associative Machinery

7. System2 == Formal Semantics

8. The Language of Thought

9. Understanding intuitive physics, the visual grammar of the world (above,
below, behind, etc.)

10. Understanding agency and causality (folk-psychology) and the necessity of
some regularities in the world

11. System2 Assumption: Compositionality

12. 𝑴 𝜶 ∘ 𝜷 = 𝒇(𝑴 𝜶 , 𝑴 𝜷 )

13. DOG PSYCHIATRIST

14. PROMPT ENGINEER

15. A horse riding another horse.
A unicorn riding an astronaut.

16. No compositionality
hallucinations stay

18. No System 2 Reasoning 
No Artificial General Intelligence

19. Face it:
Artificial General Intelligence is nowhere
near

20. On the nature of revolution
Achieved System1 Superintelligence

21. On the nature of revolution
Artificial General Intelligence  NULL

22. Achieving System1 Superintelligence
is super useful

23. Consequences

24. Client
Knowledge Graphs Knowledge Vectors
Promp Engineering
Large Language Models
Internet Search External Sources

25. AIaaS|DevsAI

26. AIaaS
value propostion:
scale
Infra cost
Training cost
Data cost
Inference cost

27. DevsAI
value propostion:
specialization
Infra cost
Compute cost
Knowledge Engineering cost

28. Strategies

29. AIaaS Strategy:
Train specialized
LLMs
Infra cost
Training cost
Data costs
Inference cost

30. How many domain-specific
models can be trained at
what respective risk?

31. BloombergGPT
• Trained on AWS SageMaker
• ~700 billion tokens dataset, only 50 billion parameters
• 64 x p4d.24xlarge instances  64 x 8 Nvidia 40GB A100 GPUs  512
GPUs total
• ~$33 per instance per hour  $2112 per hour for complete setup
• Trained for 53 days  ~$2.7 million (~$1 million with spot-pricing)

32. DevsAI
value propostion:
specialization
Infra cost
Compute cost
Knowledge Engineering costs

33. Training a general purpose
LLM is OK.

34. If you are OpenAI, Microsoft,
Google, xAI, Anthropic,
Amazon, or somewhere
near.

35. Training a specialized LLM
is risky:
DevsAI might beat you.

36. Trends

37. Not too many domain-specific LLMs:
BloombergGPT, Med-PaLM2, ClimateBERT,
BioBERT, KAI-GPT, FinGPT…

38. On the other hand:
• OpenAI introduced GPT-4 Turbo with 128K context,
• JSON response format,
• Reproducible outputs,
• Stateful API (i.e. conversation memory)…
• … and Anthropic Claude 2.1 now has a 200K context window, search and
retrieval capabilities over a variety of knowledge bases (Elasticsearch, vector
databases, web search, Wikipedia), tool use, etc.
 AIaaS is targeting the DevsAI business

39. Risks

40. AIaaS
(high cost) Infra
(high cost) Training
(high cost) Data
(high cost) Inference
DevsAI
Infra (low cost)
Knowledge (moderate cost)
Compute (high cost) ?
The emerging market risk is in the P(match):

41. DECIPHERING THE
AI LANDSCAPE:
BUSINESS, MARKET
TRENDS, AND
EMERGING RISKS