So how exactly can the screenreader interpret an image file when there is no alternative text? Even if by some extremely resource-intensive image analysis the program could identify the widget, how could it know the precise model number, size, color...?
Adding alternative text is easy - blind users will benefit, sighted readers with images turned off will benefit, if the image link is broken the text will be displayed instead (meaning your page will still convey the information even when the image doesn't load). This most certainly isn't the screenreaders fault.
No problem - try the demo version of JAWS: