What are the key challenges facing AI researchers and developers today, and how can they be addressed?

12.8k viewscircle icon2 Upvotescircle icon5 Comments
Sort by:
Data Scientist4 months ago

One of the key challenges AI researchers and developers face today is ensuring model reliability and fairness, especially in complex, real-world applications. Bias in training data, lack of transparency in decision-making (black-box models), and scalability issues also remain major concerns. These can be addressed by developing better evaluation frameworks, investing in explainable AI, diversifying datasets, and fostering interdisciplinary collaboration between ethicists, engineers, and domain experts.

Lightbulb on1
Senior Data Scientist in Miscellaneous2 years ago

I'd still regard the understanding of AI outside of the experts as the most challenging task. Everything seems to be AI now, which equals, everything that I didn't understand in math or statistics must be AI.
That's the reason, why we're all still struggling with quality and quantity for data, as the relationship is still not understood. In addition, the expected overdrawn accuracy includes the potential to reject the AI potential.
A.t.m.h.o trying to educate people continuously inside of the company and outside remains the biggest challenge to leverage overdrawn expectations to a reasonable set.

Lightbulb on3 circle icon1 Reply
no title2 years ago

The business needs to have a problem that it needs solving or an objective that it needs development for. Then the AI experts can say whether they can deliver or not. the business then needs to evaluate all options / costs etc to decide which is the best option for them. Initially the business needs to understand what they are asking for and then how AI could resolve.

Director of Digital Revenue and Marketing in Healthcare and Biotech2 years ago

Considering nothing about AI is neither artificial, nor intelligent (quoting Kate Crawford), the biggest challenge is quality and accuracy of input data. The entire efficiency AI's output depends on this.
Considering extremely fast development of this subject in the last year, it is somewhat unregulated when it comes to training data, experts, privacy, data storage, and freqauency of data updates required.
However, this topic is getting an increased attention to facilitate trust and security of AI's users in the future.

Lightbulb on1
Data Science & AI Expert in Miscellaneous2 years ago

Challenges like quantity and quality of training data is more on top of minds but here is one that might not be mentioned as frequently: Not enough familiarity with the existing regulations and uncertain up coming changes in this area. We need to raise the awareness about the different regulatory complexity level in AI systems vs conventional IT as well as closer collaboration between legal/ethics SMEs and AI experts.

Content you might like

Exploring – We are aware of SAP Joule but have not yet started any formal evaluation or planning.20%

Evaluating – We are actively assessing SAP Joule’s capabilities and fit for our business needs.47%

Piloting – We have initiated a limited rollout or proof of concept in selected areas.10%

Scaling – We are expanding SAP Joule adoption across multiple business units, processes or systems.16%

Operationalized – SAP Joule is fully integrated into our operations and delivering measurable value.6%

View Results

Unrealistic57%

Realistic34%

They aren't thinking about it right now9%

View Results