//Artefact:ServiceWebSite/com/sphenon/products/fairyloom - Introduction - Welcome/Robot
Introduction
fairyloom, Machine Learning, and AI
Manual Software Production, AI assisted
4.
AI Assistance Systems support software developers in searching and
completing coding tasks. Compared to conventional, supporting sources
such as search engines, collaboration or question-and-answer platforms,
meaningful, comprehensive answers adapted to the target problem can be
found more quickly.
Semi- or Pseudo-Automated Software Production, AI assisted
Fueled by the impressive abilities of
LLMs
for certain tasks and spurred on by the immense exploitable market volume
a nervous race for the best software production (support) tool was kicked off.
As of now, it is too early to sort all the approaches and to
separate the wheat from the chaff.
However, the assistance systems are currently reaching certain limits:
- it is not possible to understand how the system arrived at its answer
- the answers may contain errors that are difficult to recognize and must therefore be carefully checked for correctness
- the accuracy of the results decreases rapidly with the size of the task to be processed
- there is currently hardly any training material for large systems, as an exponential amount of data would quickly be required due to the large number of parameters
- with increasing size, legal risks that are difficult to recognize, such as copyright and patent infringements, can creep in
- AI systems require orders of magnitude more resources than the operation of the applicationloom
In our view, the associative-generative performance of LLMs can unfold most
effectively when embedded in a coordinating, structuring framework
that also enables the urgently needed interaction with humans.
fairyloom's
approach to this is outlined in the next section.
Production with fairyloom, AI assisted
6.
As with manual software production, AI Assistance Systems can support
software developers in searching for and completing coding tasks,
but in this case to define procedures in the
codingcookbook.
These can be checked and tested in isolation and only then
released for reproducibly reliable use.
7.
AI assistance systems can also support the editing of the
lorescroll.
Based on free-text queries similar to chatbots, the systems
suggest comprehensible changes that can be checked and,
above all, easily understood so that they can be incorporated
into the description in the
lorescroll
in a way that is accessible to both humans and machines.
AI-assisted software production with
fairyloom
effectively solves the problems described above.
We are currently testing the generation of the required training data
with the help of models. Our parameterization is already very extensive,
but more domain oriented material is needed than is currently available.
Since upcoming fairyloom
AI services will operate on one and the same
lorescroll,
your invested time, work and knowledge are protected, the scroll will
stay to be the core asset. Working with fairyloom
will just become even more easier.