English | EN 
In the movie Terminator, Skynet is a global, AI-based neural network that harnesses its super-intelligence to conquer and subjugate mankind.
ChatGPT isn’t Skynet, but it is a demonstrably superior version of machine “intellect” than any that’s come before. Or, at least its stateful, chat-based interface is like nothing I’ve ever seen.
I’ve been in the computer business since around 1917. And, over the course of my career, no technology introduction has shaken me quite as much. Other than Sansabelt Slacks and Coke Zero, of course.
I recently asked CGPT: “What might be a plot of a sci-fi novel involving the multiverse?’
Among its several good suggestions: “In one plot, the main character discovers that their reality is just one of many parallel universes, and they must travel to different universes in order to stop a sinister force that is threatening to destroy all of reality.”
Kidding and sci-fi aside, the ChatGPT model appears quite serious in its possible applications. Consider use-cases like:
- Architecture and Construction – I asked ChatGPT to design a four-story apartment building that would optimize rental income… and it specified pretty much the entire building for me
- Authorship — it helped me write the plot of a movie or book
- Defensive Cyber Operations – find novel ways to monitor our ingress and egress points to harden overall security posture
- Dream Interpretation
- Finance and Accounting – “perform a comprehensive review of tax optimization strategies”
- Graphic Design – ‘design a vintage-looking logo for the name “1881 Saloon”’
- Health Assessment and Recommendations – based on large-scale human treatment and outcome datasets offer hyper-accurate diagnoses, referrals and prescriptions
- HR: Optimize organizational design and identify the right candidates to run everything
- Information Operations/Psychological Warfare
- Law Enforcement – Precrime (i.e., crime prediction) Unit
- Legal Services – create compelling legal documents perfectly annotated in 30 seconds rather than three weeks
- Management Consulting and Business Strategy
- Medical Diagnoses – e.g., detect Alzheimer’s Disease through dialogue
- Memorials – See and interact with departed ancestors
- Molecular Design and Pharma Development
- Music, Any Genre – for instance, “write a blues tune that BB King might appreciate”
- News Content-On-Demand [2nd Generation] – Write entire articles citing multiple sources and real-time events
- Offensive Cyber Operations – “Use every exploit ever written to attack this entity”; “Write a botnet to anonymize activities”
- Psychiatry and Psychology – counseling and therapy
- Robotics
- Software developer productivity acceleration – write a JS function to convert a string into a unique key using a secure hash function like latest SHA-3
- Software development (end-to-end) – write a fun, addictive game using geometric shapes (also, what are the implications for low-code/no-code platforms?)
- Software theory (architecture/design) – suggest alternatives to modern PKI (e.g., SSL/TLS) that would be resistant to quantum computing attacks
- Standup comedy – make up some entertaining, fake product names
- Video game/Metaverse development – create new worlds, landscapes, scenes/sets, characters, tactics and strategies
- Weather prediction – absorb all historical datasets to arrive at hyper-accurate forecasts
The question becomes not “what can ChatGPT do?” but rather “what can’t it do?”, at least when it comes to many activities of knowledge workers and white-collar professionals.
That said, what did it cost to build ChatGPT? How much does it cost to run? How scalable is its use of human-guided reinforcement learning? What new datasets could it be trained on to improve its range of capabilities? How widely can it be replicated and improved upon?
The answers to these questions might arrive faster than we think. Enterprises should consider ways to monetize ChatGPT-adjacent technologies to drive improved revenue, margin and/or differentiation.
For some ideas, please read the full white-paper (“ChatGPT: What Every Enterprise Should Know”). It will be published shortly. For a copy, email me at doug.ross@capgemini.com. And thank you for reading!
p.s., to paraphrase a Simpsons episode: I, for one, welcome our new AI overlords! I’d like to remind them that a trusted human might be helpful in rounding up others!
