Pt1. Again, not sure this is useful, and has been tried many times with no useful end. The major issue is that simple summaries assume papers to be true. Just aggregating papers and making summaries is basically like search and reading abstracts. Understanding the true differences when within the morass is the hard part and usually is often very subtle.
Pt2. A lot of people talk about automation, but biology has changed so fast in the last decade that any serious automation efforts in this space have become quickly outdated. It's useful when you are doing the same thing a million times, but such things are often already automated.
I don't disagree with that, but I do think there's massive improvement waiting to be done on scientific search engines. Not sure if there will be any big breakthrough until computer can read and understand the content of scientific papers to a much higher extent than today though.
As for automation, I think that flexible automation solution shouldn't become outdated so fast. I think automating a lab to a large extent would be the same as automating a kitchen. You need to take thing in and out of storage, in a lab this would usually be different types of freezers, refrigerators, incubators etc., then you do stuff like mixing, slicing, dicing, heating, cooling, shaking, transferring liquids and solids between different types of equipment.
Smaller labs often don't have much automation. There's plenty of labs that don't even have lab robots, even though they do plenty of pipetting. Automated storage solutions are also rare as far as I know. Just deploying existing solutions and making them cheaper would help.
The only way to make thing really programmable though, is to automate everything. One challenge is that most machines and equipment only have interfaces for humans. If you were to place all your equipment in a kind of rack you could have a mix of conveyor belt for moving materials and robot arms on tracks in the roof to do manipulation tasks. Still it would be challenging to operate much of the equipment unless it is rebuilt to have standardized interfaces, but perhaps improvement in computer vision etc could eventually mitigate this. Another challenge is to make lab equipment communicate and compare data in completely different formats etc. There's lots of challenges but building a lab robotics system that can completely replace humans in doing the lab work should be possible even without hard AI and the hardware wouldn't necessarily have to be extremely expensive either.
Pt2. A lot of people talk about automation, but biology has changed so fast in the last decade that any serious automation efforts in this space have become quickly outdated. It's useful when you are doing the same thing a million times, but such things are often already automated.