AI readiness report – Scale
Developing AI projects come with many challenges, regardless of project size or industry. These challenges are within the whole AI lifecycle. If solved, the changes could yield better returns. Scale have produced its first state of AI readiness report for 2022 which outlines the challenges and solutions.
This research behind the report was based on 1,300 ML professionals across different company sizes or industry sectors. The bulk of the report looked at the issues as well as highlighting several solutions.
- 31% – said that data collection is an issue
- 33% – said that data quality is the difficult part of acquiring data
- 37% – said they do not have the variety of data
- 67% – said that data noise was a problem
- 47% – say that data bias was a problem
- 47% – said that domain gaps (knowledge) was a problem
- 9% – indicated that their data was free from noise
- 30% – said that curating data was an issues
- 33% – said that annotation quality was an issues
- 81% – said that their ML teams are somewhat or closely integrated with their annotation partners
- 73% – used synthetic data for their projects
- 51% – used synthetic data because of insufficient real-world data
- 29% – used synthetic data to address privacy or legal issues
- 28% – used synthetic data to reduce long lead times
- 80% – evaluated model performance
- 56% – used aggregated model metrics
- 51% – look at evaluating business impact
- 40% – of smaller companies (500 employees) look at impact on business
- 61% – of large companies (10,000 employees) look at business impact
- 38% – said deploying was the most challenging part
- 34% – said monitoring was difficult
- 30% – said optimizing compute was difficult
- Lots more….
Key takeaway and insights
The AI readiness report has pointed out many issues that will help improve future AI/ML development. One glaring issue is the low figure of companies measuring the impact of their models on their enterprise. It is good to have clear objectives and measurables regardless of business size. One common thread in this report is the issue of data and how the correct amount, quality and management is very important for downstream ML success.
What is needed:
- Have clear objectives for your AI and measurable towards benefits for business
- Concentrate on data quality, issues and volume
- Speeding up of curating data will help speed up deployment
- Importance of domain knowledge for labelling data
- Need for good data augmentation with fast model development, iterations and product testing
Great report, that raises some very important issues. I look forward to seeing change next year.
Read the report here (May need your details)
Purpose of this analysis
I invest many hours researching the latest AI, technology and business trends for all my clients. The knowledge is refined, distilled and combined with many other forms of analysis and knowledge in order to give the freshest and finest advice possible.
Contact me today
If you are a company in need of finding the right AI team or the right AI strategy/approach, please feel free to contact me. Whether you are start-up or multinational/government get in contact today and see how I can add value to your organisation