Artificial Intelligence is nothing new. It has been around since 1956 when John McCarthy coined the term and Alan Turing built on the concept. AI has traveled a long way but it still has miles to go despite the availability of improved software, capable AI developers and superior hardware.
1. AI is still not human
There are AI chatbots that can simulate a conversation, just like a human but AI still does not have the ability to think like a human mind. It can outfox a human player in Chess but still its intelligence and ability for fuzzy thinking is rudimentary. What this means is that when businesses (or healthcare or governments) wish to incorporate AI, they must develop goals knowing just what AI can do and cannot do. It can, for instance, analyze big data and even make recommendations to executives on possible lines of action but the executive still needs to make the right choice.
Look at it any way you want: AI costs a lot of money. It involves considerable effort, time and skill that all add to costs of development. It does require specialized hardware to leverage AI, ML and deep learning and one must also have access to large data sets for ML to learn right and let AI work right. Given this situation it is not surprising that AI is still the preserve of big businesses and governments.
3. Parallel worlds of tools
AI and ML make use of different tools and frameworks such as Spark, TensorFlow, PyTorch, H2O, Python, Google ML Kit, Theanos, Keras and MxNet to name a few. These are parallel worlds and if a developer in one area wishes to port to another it requires some effort. Each tool has its plus points and this could mean that an expert working on a solution in ML or AI should know which tool to use. That means if one engages developers for AI then one has to engage a team of experts, each with expertise in one or more of these tools and frameworks.
4. Responsibility and ethics
AI can achieve much and bright minds might see uses of AI for which it was not intended. Enterprises could go down this avenue and governments are likely to do so, given that they have so much access to citizen data. An AI developer must not only be capable but also a responsible person with a high standard of ethics because AI gives rise to a host of questions on the moral, legal and social scales.
ML is a voracious data gobbler and needs to be if it has to deliver accurate and precise results. The problem facing AI is the inadequacy of data that will be useful for training ML. If data is present it is hugely unstructured and disorganized, requiring AI to sort it all out. Acquiring data, cleaning it and making it “presentable” is just one thing. With huge data comes the need for huge storage facilities and security issues. As can be seen, the more data there is, the more AI problems are compounded.
6. Processing speeds
One of the artificial intelligence problems are that even today’s multi-core multi-threaded CPUs and GPUs cannot tackle the computationally intensive tasks in this segment. It is possible to put together CPUs in parallel and also use GPUs in parallel but the costs add up. The human brain, in its small size, has intricate neural networks. Artificial neural networks try to mimic the human brain’s behavior and to work on regression models. It is easier said than done to set up such artificial neural networks to carry out specific tasks such as study of diseases or for drug discovery.
7. Assessing capability
There are AI problems galore when it comes to the legal side. Terabytes of data are used for ML and algorithms work on such data. If such data is hacked then it makes the enterprise conducting the exercise or the developer or both liable for legal action. Then again, AI is no magic pill that will immediately replace workers and raise productivity in business. It can only assist human beings and free them from routine tasks to taking on tasks that require powers of the mind. Businesses need to know how exactly AI can help them and then assess capability of AI developer to deliver.
Even with these AI problems, the future is promising as activity in artificial intelligence progresses fast and number of actual applications increase.