dida Logo

Newsletter from April 2022

Topics: Remote Sensing Solution ASM-Spotter Receives UNESCO Award | New Case Study on NLP Language Models and Pets | What can OpenAI Codex do? | Ethics in ML projects

Dear dida friend,

after winning the "Microsoft AI for Earth Award" in 2020, our remote sensing solution ASMSpotter has now also been recognised by UNESCO's International Research Centre for Artificial Intelligence as one of ten outstanding AI solutions for achieving the Sustainable Development Goals "Responsible Consumption and Production" and "Living on Land". Developed together with our partner Levin Sources, ASMSpotter helps to detect informal mining areas in the Amazon rainforest in order to make mining more sustainable. Recently, our software has also been used by the Guyana Ministry of Natural Resources. For more information, see the Guyana Project Sheet. You can try out the tool on our ASMSpotter demo page. The technical basics of the ASMSpotter were already explained in November in our webinar "ML for Remote Sensing: Analysing Satellite Data Automatically". For more practical tips in this context, we refer you to our new blog article "The best (Python) tools for remote sensing".

We can also report on exciting new client projects in the field of NLP and ML-based language models. The company felmo helps pet lovers to find the right treatment for their pets via a website and to book a veterinary home visit directly. Using machine learning and NLP, we support felmo in automatically deriving the corresponding symptoms, diseases and desired treatments from an entered appointment reason. More details in our felmo pet case study.

How search can be improved by ML-based language models can also be tried out very practically in our new FAQ demo. For example, if you maintain FAQs for your customers, you can copy these question-answer pairs into our demo and then try asking certain questions in different ways. The language model should manage the semantic matching to the correct FAQ pair despite completely different word choices.

One particular language model is OpenAI Codex, which has caused quite a stir with its ability to generate the appropriate source code in different programming languages from the natural language description of a computer application. Our new blog article "OpenAI Codex: Why the revolution is still missing" explains why human programmers are still around.

Anyone who wants to introduce AI and especially machine learning applications can no longer avoid the topic of ethics. Our new blog article "Ethics in Natural Language Processing" clearly presents the ethical challenges that can arise, for example, from biases in the training data of the large language models described above. As always, we look forward to your feedback and any topic requests for next time. We are planning our next newsletter for the end of Q2.

With kind regards
Philipp & Lorenz

P.S.: Should you ever have more specific machine learning interests, we would be happy to offer you and your technical colleagues an exchange in which we can discuss ML content relevant to you in detail. Please feel free to contact us.