The challenge of AI and deep learning systems

One of the biggest challenges of AI is that it is not plug-n-play. You have to customize AI to your business context. A common pattern in AI is mapping an input to an output. For e.g.
  1. input a picture, output 1 if it is a cat, or 0 if it is a dog
  2. input an audio file, output transcript
  3. input Spanish text, output English text
  4. input customer logs, output 1 if customer will churn, or 0 if they will not churn
  5. input transaction logs, output 1 if transaction is fraud, or 0 if it is not
The challenge and also the opportunity of AI is discovering what input and the corresponding output that will fit in your business context.

Maybe your group writes a lot of SQL, perhaps the request for SQL comes as a ticket created in JIRA in plain english then the input in this case will be the english description of the SQL needed, along with the database schema, and the output will be SQL.

Maybe you spend time deciding if an item should be retired from your catalog. Then the input will be item details, item transaction details, customer logs and output will be 1 if retire, 0 if not retire.

Maybe you spend time deciding when to approve a return and when not to, then the input will be the return info (such as item, price, reason for return, etc) and the output will be 1 if return, 0 if not

Maybe you spend time researching 10k SEC documents for the purposes of deciding if you want to invest in this company or not. Then the input will be 10k SEC documents, and the output will be 1 if invest, and 0 if not.

Just to be clear the output does not have to binary, it can be a probability or confidence score. Also the output need not be just one class, it can be multiple classes.

As must now be clear why customizing AI is exactly what will happen over the next few years. We at SublimeAI started with exactly this belief that AI has to be customized and we are constantly searching with your help fascinating business contexts in which AI can be applied.

Many fortune 500 company CEOs have regretted that they wish they developed their internet strategy earlier, or their mobile strategy earlier. We think the same will happen with AI. Fortune 500 Company CEO's are going to regret not having developed their AI strategy in time. Building an AI strategy requires a deep understanding of AI (where we come in) and a deep understanding of your business.

We encourage you to email us at info@sublimeai.com with the business context that comes to your mind after reading this post and we would love to jump on a free consulting call to discuss if AI can indeed help.






Comments

Post a Comment

Popular posts from this blog

Large SAP tables which are good candidates for partitioning

Image
Not all tables are good candidates for partitioning. The tool we have developed runs through multiple checks for large tables to identify the right candidate. Generally if the answer is yes to the following questions, table partitioning may be a suitable strategy for improving performance. 1. Is the database large enough? 2. Is the table large enough? 3. Do your reports and queries have distinguishable access patterns? 4. Are you experiencing slowness and performance degradation? 5. Is the skew factor for the data distribution under reasonable limits? Below are some of the large tables which have been candidates for table partitioning amongst our SAP customers. Large material master tables MARA                   Material Master Table MARC                   Plant Material Data MARO...
Read more

Partitioning in SAP HANA database

SAP HANA provides two ways of distributing data 1. Database Partitioning - different tables being distributed across several hosts (SAP Note 2044468 - FAQ:SAP HANA Partitioning) 2. Table Partitioning - splitting a column store table and distributing across several hosts (SAP Note 2081591 - FAQ: SAP HANA Table Distribution)  SAP HANA supports: 1. Hash Partitioning 2. Range Partitioning 3. Round-robin Partitioning In SAP HANA, we can create 1. Single-Level Partitioning 2. Multi-Level Partitioning The optimizing algorithm takes factors into account such as: - Balance tables equally in a way that they are spread over all the slave nodes and make optimal use of the available physical memory on each node - Ensure that tables/table partitions which are belonging to the same table groups (see tables _SYS_RT.TABLE_PLACEMENT, SYS.TABLE_GROUPS) are located on the same nodes, as they are logically frequently queried from application side via joins and if t...
Read more

SAP HANA exploits partitioning in SAP BW for optimization

Image
Partitioning significantly improves query performance, based on the types of queries and on your hardware configuration as the system will find faster the requested information, because it will filter out all the irrelevant partitions and thus reduce the data volume to be read. This is a standard feature in most databases but SAP did not truly exploit this feature being database agnostic until HANA. In SAP/R3, SAP provided partition keys for BW but did not partition data during installation. With HANA the table partitioning/distribution is automated for SAP NetWeaver Business Warehouse. Database tables created for InfoCubes or classic DataStore objects are partitioned on the database. While the technical partitioning of data at database level helps optimizing the runtime of BW processes to a great extent, they alone are inadequate in building and maintaining data warehouses that scale easily with growing data volumes. Hence, to address some of the aspects of dealing with ...
Read more